chore: remove stray cd.yml.notes file from CI retrigger commit

The file was an accident in commit 24c3533 — meant as a tmp marker,
should have been removed before commit. Harmless but trash. Removing.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

.aider.conventions.md

@@ -36,6 +36,18 @@ These rules apply to every task across every project, regardless of harness.
 4. **Goal-driven execution.** Define clear success criteria up front for every task.
    Loop — implement, verify, refine — until those criteria are met. Don't claim
    completion without evidence (tests pass, command output, observed behavior).
+5. **Trunk-Based Development — commit directly to main.** Every commit is one
+   logical change (one tool, one fix, one test) with passing tests. Main is always
+   deployable. Never create long-lived feature branches.
+
+   **Exception — parallel agents on same repo:** If another agent is known to be
+   actively working on the same repo simultaneously, create a short-lived branch
+   (`agent/<description>`), finish the task, and merge to main within the same
+   session. Do not leave agent branches open between sessions.
+
+   **Exception — external contributor or client four-eyes requirement:** Use
+   PR flow only when a human reviewer outside the project is required. Document
+   the reason in PROJECT.md.
 
 ## Default stack
 
@@ -49,6 +61,7 @@ These rules apply to every task across every project, regardless of harness.
 | Search | pgvector (vector), BM25 | Qdrant (when >1M vectors or hybrid retrieval) | — |
 | Logging | slog (structured) | — | — |
 | Testing | Table-driven, testify | — | — |
+| Agents (Go) | google.golang.org/adk + pkg/litellm adapter | — | — |
 
 Exploratory: Rust, Zig — I'll tell you when I want these.
 
@@ -58,9 +71,12 @@ Exploratory: Rust, Zig — I'll tell you when I want these.
 - **Errors**: `fmt.Errorf("operation: %w", err)` — never naked, never log-and-return
 - **Naming**: stdlib conventions, no stuttering
 - **Architecture**: prefer stdlib over frameworks, constructor injection, env-var config parsed into typed structs
-- **Git**: conventional commits (`feat:`, `fix:`, `chore:`), one concern per PR, PR describes *why* not *what*
+- **Git**: conventional commits (`feat:`, `fix:`, `chore:`), commit directly to main,
+  one logical change per commit, CI is the quality gate
+- **Never**: long-lived feature branches, PRs for solo work, direct push without
+  passing `task check` locally first
 - **Security**: no secrets in code, govulncheck before adding deps, SOPS for encrypted config
-- **Dependencies**: prefer stdlib. testify, slog, templ, sqlc are pre-approved; anything else needs justification in the commit message
+- **Dependencies**: prefer stdlib. testify, slog, templ, sqlc, google.golang.org/adk (agent projects only) are pre-approved; anything else needs justification in the commit message
 
 ## Infrastructure
 
@@ -100,18 +116,64 @@ See `~/dev/PROJECT_SUMMARY.md` for detailed descriptions of each project.
 - **koala-ai-stack** (`AGENTS/`) — local AI server infrastructure management
 - **klimatkollen** (`XT/`) — Swedish municipal climate data platform
 
-## Knowledge base
+## Knowledge base — actively use it
 
-When available, agents can query the shared knowledge base:
+A persistent brain (BM25 search + LLM-synthesised Q&A) survives across sessions,
+hosts, and harnesses. It holds 100+ hard-won entries: infra incident postmortems,
+Go pitfalls, framework gotchas, design principles, ADRs. **It is not optional
+reference material — query it actively, not just when explicitly told.**
 
-- **MCP**: `mcp://hyperguild.<TAILNET>.ts.net:3100/knowledge`
-- **HTTP**: `http://hyperguild.<TAILNET>.ts.net:3100/api/v1/search`
+### When to query (treat as a reflex)
 
-<!-- TODO: replace <TAILNET> placeholder with the real Tailscale tailnet
-     name once hyperguild is deployed. Until then, agents that try to
-     reach the knowledge service on a host where it isn't running will
-     get DNS NXDOMAIN, which is the desired fail-loudly behavior. -->
-- **Scoping**: defaults to `public` collection; client projects filter to `{client}` + `public`
+- **Before** starting a non-trivial task — search for prior art with the symptom
+  AND the system component ("how did we solve X in Y?"). 5 seconds beats 5 hours.
+- **When debugging** — search for the error string, the stack frame, the affected
+  service. Past you may have already paid this tax.
+- **Before adopting** a pattern, library, framework, or model name — check if it
+  was tried and rejected, or what the integration footguns are.
+- **When making architectural decisions** — search for the domain + "ADR" or
+  "decision" to find prior reasoning before re-deriving it.
+- **When a recommendation feels novel** — challenge yourself: "has this been
+  documented?" The brain often has it.
+
+### When to write
+
+After you discover something that **future-you would forget** and that **isn't
+recoverable from the code, git log, or PR description alone**:
+
+- Bugs whose root cause is non-obvious and generalisable beyond this project.
+- Framework / library / model-name quirks that bit you and would bite anyone.
+- Design principles validated under fire (e.g. "every `_get` needs a `_list`").
+- Postmortems for incidents: what broke, why, how diagnosed, what to do next time.
+
+DON'T write project status, sprint progress, PR summaries, or "what I did this
+session" — those rot fast and the originals are in git/gitea anyway. Brain
+entries that age well are about *why*, *how to avoid*, and *what to do when*.
+
+### How to access (per harness)
+
+| Harness | Query | Write |
+|---------|-------|-------|
+| **Claude Code, Claude Desktop** | `brain_query` (BM25), `brain_answer` (LLM-synth + sources) MCP tools | `brain_write` MCP tool |
+| **Crush, Pi, Antigravity, other MCP-capable** | same MCP server: `ingestion-brain` (via the `mcp__*_brain__*` namespace once authenticated) | same |
+| **Anything HTTP-only (curl, scripts)** | `POST https://brain-mcp.d-ma.be/query` with `{"query":"..."}` (auth via `BRAIN_MCP_TOKEN`) | `POST .../write` with `{"content":"...","filename":"..."}` |
+| **Browser / human inspection** | `https://gitea.d-ma.be/mathias/hyperguild` → `knowledge/` and `wiki/` markdown files |
+
+- **Scoping**: defaults to `public` collection; client projects filter to `{client}` + `public`.
+- **Routing**: brain_answer's LLM uses berget.ai as primary, iguana ollama as
+  fallback. Both are configurable in the `supervisor/ingestion-deployment.yaml`
+  on the koala k3s cluster; don't hardcode local-only model names into the
+  berget URL (see knowledge entry on namespace mismatches).
+
+### Quick reflex checks
+
+If you find yourself about to say any of these out loud, you owe yourself a brain query first:
+
+- "I think the issue might be..."
+- "Let me try X and see..."
+- "I'll just write a script to..."
+- "This is probably a new bug..."
+- "Has anyone done this before?" — *yes, probably, go check.*
 
 ## Client work rules
 
@@ -208,9 +270,11 @@ Key skills:
 
 ### Git
 - Conventional commits: `feat:`, `fix:`, `chore:`, `docs:`, `refactor:`
-- Branch naming: `feat/short-description`, `fix/short-description`
-- PRs: one concern per PR, description explains *why* not *what*
-- **Branch protection:** always work on a feature branch, open a PR, never push directly to main
+- **Trunk-Based Development:** commit directly to main. One logical change per commit.
+- Run `task check` locally before every push. CI is the quality gate, not branch protection.
+- No feature branches, no PRs for solo/agent work.
+- Exception: if a parallel agent session is active on this repo, use a short-lived
+  `agent/<description>` branch and merge within the same session.
 
 ### Security
 - No secrets in code, ever — use env vars or SOPS-encrypted files
@@ -248,98 +312,29 @@ When acting as a coding agent on this project:
 3. If unsure about a convention, check `DECISIONS.md` or ask
 4. Never modify files outside the project root without explicit permission
 5. When adding a dependency, explain why in the commit message
-6. Always work on a feature branch and open a PR — never push directly to main
+6. Commit directly to main. Run `task check` before every push. Never create
+   feature branches unless a parallel agent is simultaneously active on this repo.
 7. For client projects: never send code or context to cloud APIs — use local models via LiteLLM
 
-## Current sprint — gitea-mcp v0.2 patch (2026-05-14)
+## Current state — v0.2.5 (2026-05-17)
 
-### Context
-The main v0.2 batch (repo_create, repo_update, repo_mirror_push, repo_delete,
-repo_tree, repo_topics_update, file_read dir-fix, issue_get, release_create,
-create_project_from_template) was implemented and pushed directly to main.
+All v0.2 work is complete and deployed. No active sprint.
 
-This sprint fixes three remaining gaps found during code review on 2026-05-14.
-These are blockers for `hyperguild new-project`.
+### What shipped
 
-### Issues to fix (all three in one PR: `fix/v02-patch`)
+| Tag | PR | Tools / fixes |
+|-----|----|---------------|
+| v0.2.2 | #21 | repo_create, repo_update, repo_mirror_push |
+| v0.2.3 | #22 | repo_tree, repo_topics_update, file_read dir fix |
+| v0.2.4 | #23 | issue_get, release_create, repo_delete |
+| v0.2.5 | #26 | repo_update archived+template, create_project_from_template template_name, pr_files_diff loop fix |
 
-#### #12 — repo_update: add `archived` and `template` fields
-**File:** `internal/gitea/repos.go` → `UpdateRepoArgs` struct
-**File:** `internal/tools/repo_update.go` → input schema + args struct
+Current main: `e31fd3f`. CI green. Deployed via Flux.
 
-Add to `UpdateRepoArgs`:
-```go
-Archived *bool
-Template *bool
-```
+### Next up
 
-Add to tool input schema:
-```json
-"archived": {
-  "type": "boolean",
-  "description": "Mark repo as archived (read-only). Requires confirm=<repo name>."
-},
-"template": {
-  "type": "boolean",
-  "description": "Toggle template repo flag."
-}
-```
+1. **`hyperguild new-project` v1** — primary next target.
+   See brain node `adr-new-project-gitea-first-github-mirror` for full flow spec.
 
-Add confirm-guard for `archived=true` (same pattern as `private=false`):
-```go
-if args.Archived != nil && *args.Archived {
-    if args.Confirm != args.Name {
-        return nil, fmt.Errorf("setting archived=true is irreversible: set confirm=%q to proceed", args.Name)
-    }
-}
-```
-
-New test cases to add in `repo_update_test.go`:
-- `TestRepoUpdateTool_Archive` — happy path with confirm
-- `TestRepoUpdateTool_ArchiveRequiresConfirm` — missing confirm returns error
-- `TestRepoUpdateTool_SetTemplate` — no confirm needed
-
-#### #24 — create_project_from_template: make template selectable
-**File:** `internal/tools/create_project_from_template.go`
-
-Add optional `template_name` param to input schema:
-```json
-"template_name": {
-  "type": "string",
-  "enum": ["template-go-web", "template-go-agent"],
-  "description": "Template repo to generate from. Defaults to template-go-web.",
-  "default": "template-go-web"
-}
-```
-
-The tool should use `args.TemplateName` if set, fall back to the hardcoded default.
-Remove the hardcoded template name from `cmd/gitea-mcp/main.go` constructor call —
-the tool resolves it internally.
-
-New test case: `TestCreateProjectFromTemplate_AgentTemplate`
-
-#### #25 — pr_files_diff: fix same diff returned for all files
-**File:** `internal/tools/pr_files_diff.go`
-
-There is a loop bug where all file entries in the response contain the same diff
-(the first file's diff is reused for every subsequent file). Find the loop and
-ensure each iteration reads and assigns the correct diff for its own file.
-
-Reproduce: call `pr_files_diff` on any PR with 3+ files, verify each file has
-a distinct diff.
-
-### Definition of done
-
-- [ ] `task check` passes
-- [ ] `repo_update` accepts `archived` and `template` params
-- [ ] `archived=true` requires `confirm=<repo name>`
-- [ ] `create_project_from_template` accepts `template_name` param, defaults to `template-go-web`
-- [ ] `pr_files_diff` returns distinct diff per file
-- [ ] All new test cases pass
-- [ ] PR `fix/v02-patch` merged to main via PR (not direct push)
-
-### After this sprint
-
-Next: `hyperguild new-project` v1 implementation.
-See brain node `adr-new-project-gitea-first-github-mirror` for the full flow spec.
-Also: verify end-to-end mirror flow (issue #19) once `repo_mirror_push` is confirmed working.
+2. **Issue #19** — end-to-end mirror flow verification.
+   `repo_mirror_push` is implemented but the full flow (create repo → add push mirror → verify sync to GitHub) has not been tested manually. Do this before relying on it in production.

.context/PROJECT.md

@@ -37,9 +37,11 @@
 
 ### Git
 - Conventional commits: `feat:`, `fix:`, `chore:`, `docs:`, `refactor:`
-- Branch naming: `feat/short-description`, `fix/short-description`
-- PRs: one concern per PR, description explains *why* not *what*
-- **Branch protection:** always work on a feature branch, open a PR, never push directly to main
+- **Trunk-Based Development:** commit directly to main. One logical change per commit.
+- Run `task check` locally before every push. CI is the quality gate, not branch protection.
+- No feature branches, no PRs for solo/agent work.
+- Exception: if a parallel agent session is active on this repo, use a short-lived
+  `agent/<description>` branch and merge within the same session.
 
 ### Security
 - No secrets in code, ever — use env vars or SOPS-encrypted files
@@ -77,98 +79,29 @@ When acting as a coding agent on this project:
 3. If unsure about a convention, check `DECISIONS.md` or ask
 4. Never modify files outside the project root without explicit permission
 5. When adding a dependency, explain why in the commit message
-6. Always work on a feature branch and open a PR — never push directly to main
+6. Commit directly to main. Run `task check` before every push. Never create
+   feature branches unless a parallel agent is simultaneously active on this repo.
 7. For client projects: never send code or context to cloud APIs — use local models via LiteLLM
 
-## Current sprint — gitea-mcp v0.2 patch (2026-05-14)
+## Current state — v0.2.5 (2026-05-17)
 
-### Context
-The main v0.2 batch (repo_create, repo_update, repo_mirror_push, repo_delete,
-repo_tree, repo_topics_update, file_read dir-fix, issue_get, release_create,
-create_project_from_template) was implemented and pushed directly to main.
+All v0.2 work is complete and deployed. No active sprint.
 
-This sprint fixes three remaining gaps found during code review on 2026-05-14.
-These are blockers for `hyperguild new-project`.
+### What shipped
 
-### Issues to fix (all three in one PR: `fix/v02-patch`)
+| Tag | PR | Tools / fixes |
+|-----|----|---------------|
+| v0.2.2 | #21 | repo_create, repo_update, repo_mirror_push |
+| v0.2.3 | #22 | repo_tree, repo_topics_update, file_read dir fix |
+| v0.2.4 | #23 | issue_get, release_create, repo_delete |
+| v0.2.5 | #26 | repo_update archived+template, create_project_from_template template_name, pr_files_diff loop fix |
 
-#### #12 — repo_update: add `archived` and `template` fields
-**File:** `internal/gitea/repos.go` → `UpdateRepoArgs` struct
-**File:** `internal/tools/repo_update.go` → input schema + args struct
+Current main: `e31fd3f`. CI green. Deployed via Flux.
 
-Add to `UpdateRepoArgs`:
-```go
-Archived *bool
-Template *bool
-```
+### Next up
 
-Add to tool input schema:
-```json
-"archived": {
-  "type": "boolean",
-  "description": "Mark repo as archived (read-only). Requires confirm=<repo name>."
-},
-"template": {
-  "type": "boolean",
-  "description": "Toggle template repo flag."
-}
-```
+1. **`hyperguild new-project` v1** — primary next target.
+   See brain node `adr-new-project-gitea-first-github-mirror` for full flow spec.
 
-Add confirm-guard for `archived=true` (same pattern as `private=false`):
-```go
-if args.Archived != nil && *args.Archived {
-    if args.Confirm != args.Name {
-        return nil, fmt.Errorf("setting archived=true is irreversible: set confirm=%q to proceed", args.Name)
-    }
-}
-```
-
-New test cases to add in `repo_update_test.go`:
-- `TestRepoUpdateTool_Archive` — happy path with confirm
-- `TestRepoUpdateTool_ArchiveRequiresConfirm` — missing confirm returns error
-- `TestRepoUpdateTool_SetTemplate` — no confirm needed
-
-#### #24 — create_project_from_template: make template selectable
-**File:** `internal/tools/create_project_from_template.go`
-
-Add optional `template_name` param to input schema:
-```json
-"template_name": {
-  "type": "string",
-  "enum": ["template-go-web", "template-go-agent"],
-  "description": "Template repo to generate from. Defaults to template-go-web.",
-  "default": "template-go-web"
-}
-```
-
-The tool should use `args.TemplateName` if set, fall back to the hardcoded default.
-Remove the hardcoded template name from `cmd/gitea-mcp/main.go` constructor call —
-the tool resolves it internally.
-
-New test case: `TestCreateProjectFromTemplate_AgentTemplate`
-
-#### #25 — pr_files_diff: fix same diff returned for all files
-**File:** `internal/tools/pr_files_diff.go`
-
-There is a loop bug where all file entries in the response contain the same diff
-(the first file's diff is reused for every subsequent file). Find the loop and
-ensure each iteration reads and assigns the correct diff for its own file.
-
-Reproduce: call `pr_files_diff` on any PR with 3+ files, verify each file has
-a distinct diff.
-
-### Definition of done
-
-- [ ] `task check` passes
-- [ ] `repo_update` accepts `archived` and `template` params
-- [ ] `archived=true` requires `confirm=<repo name>`
-- [ ] `create_project_from_template` accepts `template_name` param, defaults to `template-go-web`
-- [ ] `pr_files_diff` returns distinct diff per file
-- [ ] All new test cases pass
-- [ ] PR `fix/v02-patch` merged to main via PR (not direct push)
-
-### After this sprint
-
-Next: `hyperguild new-project` v1 implementation.
-See brain node `adr-new-project-gitea-first-github-mirror` for the full flow spec.
-Also: verify end-to-end mirror flow (issue #19) once `repo_mirror_push` is confirmed working.
+2. **Issue #19** — end-to-end mirror flow verification.
+   `repo_mirror_push` is implemented but the full flow (create repo → add push mirror → verify sync to GitHub) has not been tested manually. Do this before relying on it in production.

.context/mcp.json

@@ -16,7 +16,10 @@
     },
     "infra": {
       "type": "http",
-      "url": "https://infra-mcp.d-ma.be/mcp"
+      "url": "https://infra-mcp.d-ma.be/mcp",
+      "headers": {
+        "Authorization": "Bearer ${INFRA_MCP_TOKEN}"
+      }
     }
   }
 }

.context/system-prompt.txt

@@ -41,6 +41,18 @@ These rules apply to every task across every project, regardless of harness.
 4. **Goal-driven execution.** Define clear success criteria up front for every task.
    Loop — implement, verify, refine — until those criteria are met. Don't claim
    completion without evidence (tests pass, command output, observed behavior).
+5. **Trunk-Based Development — commit directly to main.** Every commit is one
+   logical change (one tool, one fix, one test) with passing tests. Main is always
+   deployable. Never create long-lived feature branches.
+
+   **Exception — parallel agents on same repo:** If another agent is known to be
+   actively working on the same repo simultaneously, create a short-lived branch
+   (`agent/<description>`), finish the task, and merge to main within the same
+   session. Do not leave agent branches open between sessions.
+
+   **Exception — external contributor or client four-eyes requirement:** Use
+   PR flow only when a human reviewer outside the project is required. Document
+   the reason in PROJECT.md.
 
 ## Default stack
 
@@ -54,6 +66,7 @@ These rules apply to every task across every project, regardless of harness.
 | Search | pgvector (vector), BM25 | Qdrant (when >1M vectors or hybrid retrieval) | — |
 | Logging | slog (structured) | — | — |
 | Testing | Table-driven, testify | — | — |
+| Agents (Go) | google.golang.org/adk + pkg/litellm adapter | — | — |
 
 Exploratory: Rust, Zig — I'll tell you when I want these.
 
@@ -63,9 +76,12 @@ Exploratory: Rust, Zig — I'll tell you when I want these.
 - **Errors**: `fmt.Errorf("operation: %w", err)` — never naked, never log-and-return
 - **Naming**: stdlib conventions, no stuttering
 - **Architecture**: prefer stdlib over frameworks, constructor injection, env-var config parsed into typed structs
-- **Git**: conventional commits (`feat:`, `fix:`, `chore:`), one concern per PR, PR describes *why* not *what*
+- **Git**: conventional commits (`feat:`, `fix:`, `chore:`), commit directly to main,
+  one logical change per commit, CI is the quality gate
+- **Never**: long-lived feature branches, PRs for solo work, direct push without
+  passing `task check` locally first
 - **Security**: no secrets in code, govulncheck before adding deps, SOPS for encrypted config
-- **Dependencies**: prefer stdlib. testify, slog, templ, sqlc are pre-approved; anything else needs justification in the commit message
+- **Dependencies**: prefer stdlib. testify, slog, templ, sqlc, google.golang.org/adk (agent projects only) are pre-approved; anything else needs justification in the commit message
 
 ## Infrastructure
 
@@ -105,18 +121,64 @@ See `~/dev/PROJECT_SUMMARY.md` for detailed descriptions of each project.
 - **koala-ai-stack** (`AGENTS/`) — local AI server infrastructure management
 - **klimatkollen** (`XT/`) — Swedish municipal climate data platform
 
-## Knowledge base
+## Knowledge base — actively use it
 
-When available, agents can query the shared knowledge base:
+A persistent brain (BM25 search + LLM-synthesised Q&A) survives across sessions,
+hosts, and harnesses. It holds 100+ hard-won entries: infra incident postmortems,
+Go pitfalls, framework gotchas, design principles, ADRs. **It is not optional
+reference material — query it actively, not just when explicitly told.**
 
-- **MCP**: `mcp://hyperguild.<TAILNET>.ts.net:3100/knowledge`
-- **HTTP**: `http://hyperguild.<TAILNET>.ts.net:3100/api/v1/search`
+### When to query (treat as a reflex)
 
-<!-- TODO: replace <TAILNET> placeholder with the real Tailscale tailnet
-     name once hyperguild is deployed. Until then, agents that try to
-     reach the knowledge service on a host where it isn't running will
-     get DNS NXDOMAIN, which is the desired fail-loudly behavior. -->
-- **Scoping**: defaults to `public` collection; client projects filter to `{client}` + `public`
+- **Before** starting a non-trivial task — search for prior art with the symptom
+  AND the system component ("how did we solve X in Y?"). 5 seconds beats 5 hours.
+- **When debugging** — search for the error string, the stack frame, the affected
+  service. Past you may have already paid this tax.
+- **Before adopting** a pattern, library, framework, or model name — check if it
+  was tried and rejected, or what the integration footguns are.
+- **When making architectural decisions** — search for the domain + "ADR" or
+  "decision" to find prior reasoning before re-deriving it.
+- **When a recommendation feels novel** — challenge yourself: "has this been
+  documented?" The brain often has it.
+
+### When to write
+
+After you discover something that **future-you would forget** and that **isn't
+recoverable from the code, git log, or PR description alone**:
+
+- Bugs whose root cause is non-obvious and generalisable beyond this project.
+- Framework / library / model-name quirks that bit you and would bite anyone.
+- Design principles validated under fire (e.g. "every `_get` needs a `_list`").
+- Postmortems for incidents: what broke, why, how diagnosed, what to do next time.
+
+DON'T write project status, sprint progress, PR summaries, or "what I did this
+session" — those rot fast and the originals are in git/gitea anyway. Brain
+entries that age well are about *why*, *how to avoid*, and *what to do when*.
+
+### How to access (per harness)
+
+| Harness | Query | Write |
+|---------|-------|-------|
+| **Claude Code, Claude Desktop** | `brain_query` (BM25), `brain_answer` (LLM-synth + sources) MCP tools | `brain_write` MCP tool |
+| **Crush, Pi, Antigravity, other MCP-capable** | same MCP server: `ingestion-brain` (via the `mcp__*_brain__*` namespace once authenticated) | same |
+| **Anything HTTP-only (curl, scripts)** | `POST https://brain-mcp.d-ma.be/query` with `{"query":"..."}` (auth via `BRAIN_MCP_TOKEN`) | `POST .../write` with `{"content":"...","filename":"..."}` |
+| **Browser / human inspection** | `https://gitea.d-ma.be/mathias/hyperguild` → `knowledge/` and `wiki/` markdown files |
+
+- **Scoping**: defaults to `public` collection; client projects filter to `{client}` + `public`.
+- **Routing**: brain_answer's LLM uses berget.ai as primary, iguana ollama as
+  fallback. Both are configurable in the `supervisor/ingestion-deployment.yaml`
+  on the koala k3s cluster; don't hardcode local-only model names into the
+  berget URL (see knowledge entry on namespace mismatches).
+
+### Quick reflex checks
+
+If you find yourself about to say any of these out loud, you owe yourself a brain query first:
+
+- "I think the issue might be..."
+- "Let me try X and see..."
+- "I'll just write a script to..."
+- "This is probably a new bug..."
+- "Has anyone done this before?" — *yes, probably, go check.*
 
 ## Client work rules
 
@@ -213,9 +275,11 @@ Key skills:
 
 ### Git
 - Conventional commits: `feat:`, `fix:`, `chore:`, `docs:`, `refactor:`
-- Branch naming: `feat/short-description`, `fix/short-description`
-- PRs: one concern per PR, description explains *why* not *what*
-- **Branch protection:** always work on a feature branch, open a PR, never push directly to main
+- **Trunk-Based Development:** commit directly to main. One logical change per commit.
+- Run `task check` locally before every push. CI is the quality gate, not branch protection.
+- No feature branches, no PRs for solo/agent work.
+- Exception: if a parallel agent session is active on this repo, use a short-lived
+  `agent/<description>` branch and merge within the same session.
 
 ### Security
 - No secrets in code, ever — use env vars or SOPS-encrypted files
@@ -253,100 +317,31 @@ When acting as a coding agent on this project:
 3. If unsure about a convention, check `DECISIONS.md` or ask
 4. Never modify files outside the project root without explicit permission
 5. When adding a dependency, explain why in the commit message
-6. Always work on a feature branch and open a PR — never push directly to main
+6. Commit directly to main. Run `task check` before every push. Never create
+   feature branches unless a parallel agent is simultaneously active on this repo.
 7. For client projects: never send code or context to cloud APIs — use local models via LiteLLM
 
-## Current sprint — gitea-mcp v0.2 patch (2026-05-14)
+## Current state — v0.2.5 (2026-05-17)
 
-### Context
-The main v0.2 batch (repo_create, repo_update, repo_mirror_push, repo_delete,
-repo_tree, repo_topics_update, file_read dir-fix, issue_get, release_create,
-create_project_from_template) was implemented and pushed directly to main.
+All v0.2 work is complete and deployed. No active sprint.
 
-This sprint fixes three remaining gaps found during code review on 2026-05-14.
-These are blockers for `hyperguild new-project`.
+### What shipped
 
-### Issues to fix (all three in one PR: `fix/v02-patch`)
+| Tag | PR | Tools / fixes |
+|-----|----|---------------|
+| v0.2.2 | #21 | repo_create, repo_update, repo_mirror_push |
+| v0.2.3 | #22 | repo_tree, repo_topics_update, file_read dir fix |
+| v0.2.4 | #23 | issue_get, release_create, repo_delete |
+| v0.2.5 | #26 | repo_update archived+template, create_project_from_template template_name, pr_files_diff loop fix |
 
-#### #12 — repo_update: add `archived` and `template` fields
-**File:** `internal/gitea/repos.go` → `UpdateRepoArgs` struct
-**File:** `internal/tools/repo_update.go` → input schema + args struct
+Current main: `e31fd3f`. CI green. Deployed via Flux.
 
-Add to `UpdateRepoArgs`:
-```go
-Archived *bool
-Template *bool
-```
+### Next up
 
-Add to tool input schema:
-```json
-"archived": {
-  "type": "boolean",
-  "description": "Mark repo as archived (read-only). Requires confirm=<repo name>."
-},
-"template": {
-  "type": "boolean",
-  "description": "Toggle template repo flag."
-}
-```
+1. **`hyperguild new-project` v1** — primary next target.
+   See brain node `adr-new-project-gitea-first-github-mirror` for full flow spec.
 
-Add confirm-guard for `archived=true` (same pattern as `private=false`):
-```go
-if args.Archived != nil && *args.Archived {
-    if args.Confirm != args.Name {
-        return nil, fmt.Errorf("setting archived=true is irreversible: set confirm=%q to proceed", args.Name)
-    }
-}
-```
-
-New test cases to add in `repo_update_test.go`:
-- `TestRepoUpdateTool_Archive` — happy path with confirm
-- `TestRepoUpdateTool_ArchiveRequiresConfirm` — missing confirm returns error
-- `TestRepoUpdateTool_SetTemplate` — no confirm needed
-
-#### #24 — create_project_from_template: make template selectable
-**File:** `internal/tools/create_project_from_template.go`
-
-Add optional `template_name` param to input schema:
-```json
-"template_name": {
-  "type": "string",
-  "enum": ["template-go-web", "template-go-agent"],
-  "description": "Template repo to generate from. Defaults to template-go-web.",
-  "default": "template-go-web"
-}
-```
-
-The tool should use `args.TemplateName` if set, fall back to the hardcoded default.
-Remove the hardcoded template name from `cmd/gitea-mcp/main.go` constructor call —
-the tool resolves it internally.
-
-New test case: `TestCreateProjectFromTemplate_AgentTemplate`
-
-#### #25 — pr_files_diff: fix same diff returned for all files
-**File:** `internal/tools/pr_files_diff.go`
-
-There is a loop bug where all file entries in the response contain the same diff
-(the first file's diff is reused for every subsequent file). Find the loop and
-ensure each iteration reads and assigns the correct diff for its own file.
-
-Reproduce: call `pr_files_diff` on any PR with 3+ files, verify each file has
-a distinct diff.
-
-### Definition of done
-
-- [ ] `task check` passes
-- [ ] `repo_update` accepts `archived` and `template` params
-- [ ] `archived=true` requires `confirm=<repo name>`
-- [ ] `create_project_from_template` accepts `template_name` param, defaults to `template-go-web`
-- [ ] `pr_files_diff` returns distinct diff per file
-- [ ] All new test cases pass
-- [ ] PR `fix/v02-patch` merged to main via PR (not direct push)
-
-### After this sprint
-
-Next: `hyperguild new-project` v1 implementation.
-See brain node `adr-new-project-gitea-first-github-mirror` for the full flow spec.
-Also: verify end-to-end mirror flow (issue #19) once `repo_mirror_push` is confirmed working.
+2. **Issue #19** — end-to-end mirror flow verification.
+   `repo_mirror_push` is implemented but the full flow (create repo → add push mirror → verify sync to GitHub) has not been tested manually. Do this before relying on it in production.
 
 ---

.cursorrules

@@ -39,6 +39,18 @@ These rules apply to every task across every project, regardless of harness.
 4. **Goal-driven execution.** Define clear success criteria up front for every task.
    Loop — implement, verify, refine — until those criteria are met. Don't claim
    completion without evidence (tests pass, command output, observed behavior).
+5. **Trunk-Based Development — commit directly to main.** Every commit is one
+   logical change (one tool, one fix, one test) with passing tests. Main is always
+   deployable. Never create long-lived feature branches.
+
+   **Exception — parallel agents on same repo:** If another agent is known to be
+   actively working on the same repo simultaneously, create a short-lived branch
+   (`agent/<description>`), finish the task, and merge to main within the same
+   session. Do not leave agent branches open between sessions.
+
+   **Exception — external contributor or client four-eyes requirement:** Use
+   PR flow only when a human reviewer outside the project is required. Document
+   the reason in PROJECT.md.
 
 ## Default stack
 
@@ -52,6 +64,7 @@ These rules apply to every task across every project, regardless of harness.
 | Search | pgvector (vector), BM25 | Qdrant (when >1M vectors or hybrid retrieval) | — |
 | Logging | slog (structured) | — | — |
 | Testing | Table-driven, testify | — | — |
+| Agents (Go) | google.golang.org/adk + pkg/litellm adapter | — | — |
 
 Exploratory: Rust, Zig — I'll tell you when I want these.
 
@@ -61,9 +74,12 @@ Exploratory: Rust, Zig — I'll tell you when I want these.
 - **Errors**: `fmt.Errorf("operation: %w", err)` — never naked, never log-and-return
 - **Naming**: stdlib conventions, no stuttering
 - **Architecture**: prefer stdlib over frameworks, constructor injection, env-var config parsed into typed structs
-- **Git**: conventional commits (`feat:`, `fix:`, `chore:`), one concern per PR, PR describes *why* not *what*
+- **Git**: conventional commits (`feat:`, `fix:`, `chore:`), commit directly to main,
+  one logical change per commit, CI is the quality gate
+- **Never**: long-lived feature branches, PRs for solo work, direct push without
+  passing `task check` locally first
 - **Security**: no secrets in code, govulncheck before adding deps, SOPS for encrypted config
-- **Dependencies**: prefer stdlib. testify, slog, templ, sqlc are pre-approved; anything else needs justification in the commit message
+- **Dependencies**: prefer stdlib. testify, slog, templ, sqlc, google.golang.org/adk (agent projects only) are pre-approved; anything else needs justification in the commit message
 
 ## Infrastructure
 
@@ -103,18 +119,64 @@ See `~/dev/PROJECT_SUMMARY.md` for detailed descriptions of each project.
 - **koala-ai-stack** (`AGENTS/`) — local AI server infrastructure management
 - **klimatkollen** (`XT/`) — Swedish municipal climate data platform
 
-## Knowledge base
+## Knowledge base — actively use it
 
-When available, agents can query the shared knowledge base:
+A persistent brain (BM25 search + LLM-synthesised Q&A) survives across sessions,
+hosts, and harnesses. It holds 100+ hard-won entries: infra incident postmortems,
+Go pitfalls, framework gotchas, design principles, ADRs. **It is not optional
+reference material — query it actively, not just when explicitly told.**
 
-- **MCP**: `mcp://hyperguild.<TAILNET>.ts.net:3100/knowledge`
-- **HTTP**: `http://hyperguild.<TAILNET>.ts.net:3100/api/v1/search`
+### When to query (treat as a reflex)
 
-<!-- TODO: replace <TAILNET> placeholder with the real Tailscale tailnet
-     name once hyperguild is deployed. Until then, agents that try to
-     reach the knowledge service on a host where it isn't running will
-     get DNS NXDOMAIN, which is the desired fail-loudly behavior. -->
-- **Scoping**: defaults to `public` collection; client projects filter to `{client}` + `public`
+- **Before** starting a non-trivial task — search for prior art with the symptom
+  AND the system component ("how did we solve X in Y?"). 5 seconds beats 5 hours.
+- **When debugging** — search for the error string, the stack frame, the affected
+  service. Past you may have already paid this tax.
+- **Before adopting** a pattern, library, framework, or model name — check if it
+  was tried and rejected, or what the integration footguns are.
+- **When making architectural decisions** — search for the domain + "ADR" or
+  "decision" to find prior reasoning before re-deriving it.
+- **When a recommendation feels novel** — challenge yourself: "has this been
+  documented?" The brain often has it.
+
+### When to write
+
+After you discover something that **future-you would forget** and that **isn't
+recoverable from the code, git log, or PR description alone**:
+
+- Bugs whose root cause is non-obvious and generalisable beyond this project.
+- Framework / library / model-name quirks that bit you and would bite anyone.
+- Design principles validated under fire (e.g. "every `_get` needs a `_list`").
+- Postmortems for incidents: what broke, why, how diagnosed, what to do next time.
+
+DON'T write project status, sprint progress, PR summaries, or "what I did this
+session" — those rot fast and the originals are in git/gitea anyway. Brain
+entries that age well are about *why*, *how to avoid*, and *what to do when*.
+
+### How to access (per harness)
+
+| Harness | Query | Write |
+|---------|-------|-------|
+| **Claude Code, Claude Desktop** | `brain_query` (BM25), `brain_answer` (LLM-synth + sources) MCP tools | `brain_write` MCP tool |
+| **Crush, Pi, Antigravity, other MCP-capable** | same MCP server: `ingestion-brain` (via the `mcp__*_brain__*` namespace once authenticated) | same |
+| **Anything HTTP-only (curl, scripts)** | `POST https://brain-mcp.d-ma.be/query` with `{"query":"..."}` (auth via `BRAIN_MCP_TOKEN`) | `POST .../write` with `{"content":"...","filename":"..."}` |
+| **Browser / human inspection** | `https://gitea.d-ma.be/mathias/hyperguild` → `knowledge/` and `wiki/` markdown files |
+
+- **Scoping**: defaults to `public` collection; client projects filter to `{client}` + `public`.
+- **Routing**: brain_answer's LLM uses berget.ai as primary, iguana ollama as
+  fallback. Both are configurable in the `supervisor/ingestion-deployment.yaml`
+  on the koala k3s cluster; don't hardcode local-only model names into the
+  berget URL (see knowledge entry on namespace mismatches).
+
+### Quick reflex checks
+
+If you find yourself about to say any of these out loud, you owe yourself a brain query first:
+
+- "I think the issue might be..."
+- "Let me try X and see..."
+- "I'll just write a script to..."
+- "This is probably a new bug..."
+- "Has anyone done this before?" — *yes, probably, go check.*
 
 ## Client work rules
 
@@ -211,9 +273,11 @@ Key skills:
 
 ### Git
 - Conventional commits: `feat:`, `fix:`, `chore:`, `docs:`, `refactor:`
-- Branch naming: `feat/short-description`, `fix/short-description`
-- PRs: one concern per PR, description explains *why* not *what*
-- **Branch protection:** always work on a feature branch, open a PR, never push directly to main
+- **Trunk-Based Development:** commit directly to main. One logical change per commit.
+- Run `task check` locally before every push. CI is the quality gate, not branch protection.
+- No feature branches, no PRs for solo/agent work.
+- Exception: if a parallel agent session is active on this repo, use a short-lived
+  `agent/<description>` branch and merge within the same session.
 
 ### Security
 - No secrets in code, ever — use env vars or SOPS-encrypted files
@@ -251,98 +315,29 @@ When acting as a coding agent on this project:
 3. If unsure about a convention, check `DECISIONS.md` or ask
 4. Never modify files outside the project root without explicit permission
 5. When adding a dependency, explain why in the commit message
-6. Always work on a feature branch and open a PR — never push directly to main
+6. Commit directly to main. Run `task check` before every push. Never create
+   feature branches unless a parallel agent is simultaneously active on this repo.
 7. For client projects: never send code or context to cloud APIs — use local models via LiteLLM
 
-## Current sprint — gitea-mcp v0.2 patch (2026-05-14)
+## Current state — v0.2.5 (2026-05-17)
 
-### Context
-The main v0.2 batch (repo_create, repo_update, repo_mirror_push, repo_delete,
-repo_tree, repo_topics_update, file_read dir-fix, issue_get, release_create,
-create_project_from_template) was implemented and pushed directly to main.
+All v0.2 work is complete and deployed. No active sprint.
 
-This sprint fixes three remaining gaps found during code review on 2026-05-14.
-These are blockers for `hyperguild new-project`.
+### What shipped
 
-### Issues to fix (all three in one PR: `fix/v02-patch`)
+| Tag | PR | Tools / fixes |
+|-----|----|---------------|
+| v0.2.2 | #21 | repo_create, repo_update, repo_mirror_push |
+| v0.2.3 | #22 | repo_tree, repo_topics_update, file_read dir fix |
+| v0.2.4 | #23 | issue_get, release_create, repo_delete |
+| v0.2.5 | #26 | repo_update archived+template, create_project_from_template template_name, pr_files_diff loop fix |
 
-#### #12 — repo_update: add `archived` and `template` fields
-**File:** `internal/gitea/repos.go` → `UpdateRepoArgs` struct
-**File:** `internal/tools/repo_update.go` → input schema + args struct
+Current main: `e31fd3f`. CI green. Deployed via Flux.
 
-Add to `UpdateRepoArgs`:
-```go
-Archived *bool
-Template *bool
-```
+### Next up
 
-Add to tool input schema:
-```json
-"archived": {
-  "type": "boolean",
-  "description": "Mark repo as archived (read-only). Requires confirm=<repo name>."
-},
-"template": {
-  "type": "boolean",
-  "description": "Toggle template repo flag."
-}
-```
+1. **`hyperguild new-project` v1** — primary next target.
+   See brain node `adr-new-project-gitea-first-github-mirror` for full flow spec.
 
-Add confirm-guard for `archived=true` (same pattern as `private=false`):
-```go
-if args.Archived != nil && *args.Archived {
-    if args.Confirm != args.Name {
-        return nil, fmt.Errorf("setting archived=true is irreversible: set confirm=%q to proceed", args.Name)
-    }
-}
-```
-
-New test cases to add in `repo_update_test.go`:
-- `TestRepoUpdateTool_Archive` — happy path with confirm
-- `TestRepoUpdateTool_ArchiveRequiresConfirm` — missing confirm returns error
-- `TestRepoUpdateTool_SetTemplate` — no confirm needed
-
-#### #24 — create_project_from_template: make template selectable
-**File:** `internal/tools/create_project_from_template.go`
-
-Add optional `template_name` param to input schema:
-```json
-"template_name": {
-  "type": "string",
-  "enum": ["template-go-web", "template-go-agent"],
-  "description": "Template repo to generate from. Defaults to template-go-web.",
-  "default": "template-go-web"
-}
-```
-
-The tool should use `args.TemplateName` if set, fall back to the hardcoded default.
-Remove the hardcoded template name from `cmd/gitea-mcp/main.go` constructor call —
-the tool resolves it internally.
-
-New test case: `TestCreateProjectFromTemplate_AgentTemplate`
-
-#### #25 — pr_files_diff: fix same diff returned for all files
-**File:** `internal/tools/pr_files_diff.go`
-
-There is a loop bug where all file entries in the response contain the same diff
-(the first file's diff is reused for every subsequent file). Find the loop and
-ensure each iteration reads and assigns the correct diff for its own file.
-
-Reproduce: call `pr_files_diff` on any PR with 3+ files, verify each file has
-a distinct diff.
-
-### Definition of done
-
-- [ ] `task check` passes
-- [ ] `repo_update` accepts `archived` and `template` params
-- [ ] `archived=true` requires `confirm=<repo name>`
-- [ ] `create_project_from_template` accepts `template_name` param, defaults to `template-go-web`
-- [ ] `pr_files_diff` returns distinct diff per file
-- [ ] All new test cases pass
-- [ ] PR `fix/v02-patch` merged to main via PR (not direct push)
-
-### After this sprint
-
-Next: `hyperguild new-project` v1 implementation.
-See brain node `adr-new-project-gitea-first-github-mirror` for the full flow spec.
-Also: verify end-to-end mirror flow (issue #19) once `repo_mirror_push` is confirmed working.
+2. **Issue #19** — end-to-end mirror flow verification.
+   `repo_mirror_push` is implemented but the full flow (create repo → add push mirror → verify sync to GitHub) has not been tested manually. Do this before relying on it in production.

.gitea/workflows/cd.yml

@@ -4,8 +4,6 @@ on:
   push:
     branches: [main]
     tags: ["v*"]
-  pull_request:
-    branches: [main]
 
 env:
   IMAGE: gitea-mcp
@@ -43,7 +41,6 @@ jobs:
     name: Build & Import
     needs: check
     runs-on: self-hosted
-    if: github.event_name != 'pull_request'
     outputs:
       image-tag: ${{ steps.meta.outputs.sha-tag }}
     steps:

.githooks/pre-push

@@ -0,0 +1,5 @@
+#!/usr/bin/env bash
+set -euo pipefail
+echo "→ Running task check before push..."
+task check
+echo "✓ pre-push check passed"

AGENTS.md

@@ -36,6 +36,18 @@ These rules apply to every task across every project, regardless of harness.
 4. **Goal-driven execution.** Define clear success criteria up front for every task.
    Loop — implement, verify, refine — until those criteria are met. Don't claim
    completion without evidence (tests pass, command output, observed behavior).
+5. **Trunk-Based Development — commit directly to main.** Every commit is one
+   logical change (one tool, one fix, one test) with passing tests. Main is always
+   deployable. Never create long-lived feature branches.
+
+   **Exception — parallel agents on same repo:** If another agent is known to be
+   actively working on the same repo simultaneously, create a short-lived branch
+   (`agent/<description>`), finish the task, and merge to main within the same
+   session. Do not leave agent branches open between sessions.
+
+   **Exception — external contributor or client four-eyes requirement:** Use
+   PR flow only when a human reviewer outside the project is required. Document
+   the reason in PROJECT.md.
 
 ## Default stack
 
@@ -49,6 +61,7 @@ These rules apply to every task across every project, regardless of harness.
 | Search | pgvector (vector), BM25 | Qdrant (when >1M vectors or hybrid retrieval) | — |
 | Logging | slog (structured) | — | — |
 | Testing | Table-driven, testify | — | — |
+| Agents (Go) | google.golang.org/adk + pkg/litellm adapter | — | — |
 
 Exploratory: Rust, Zig — I'll tell you when I want these.
 
@@ -58,9 +71,12 @@ Exploratory: Rust, Zig — I'll tell you when I want these.
 - **Errors**: `fmt.Errorf("operation: %w", err)` — never naked, never log-and-return
 - **Naming**: stdlib conventions, no stuttering
 - **Architecture**: prefer stdlib over frameworks, constructor injection, env-var config parsed into typed structs
-- **Git**: conventional commits (`feat:`, `fix:`, `chore:`), one concern per PR, PR describes *why* not *what*
+- **Git**: conventional commits (`feat:`, `fix:`, `chore:`), commit directly to main,
+  one logical change per commit, CI is the quality gate
+- **Never**: long-lived feature branches, PRs for solo work, direct push without
+  passing `task check` locally first
 - **Security**: no secrets in code, govulncheck before adding deps, SOPS for encrypted config
-- **Dependencies**: prefer stdlib. testify, slog, templ, sqlc are pre-approved; anything else needs justification in the commit message
+- **Dependencies**: prefer stdlib. testify, slog, templ, sqlc, google.golang.org/adk (agent projects only) are pre-approved; anything else needs justification in the commit message
 
 ## Infrastructure
 
@@ -100,18 +116,64 @@ See `~/dev/PROJECT_SUMMARY.md` for detailed descriptions of each project.
 - **koala-ai-stack** (`AGENTS/`) — local AI server infrastructure management
 - **klimatkollen** (`XT/`) — Swedish municipal climate data platform
 
-## Knowledge base
+## Knowledge base — actively use it
 
-When available, agents can query the shared knowledge base:
+A persistent brain (BM25 search + LLM-synthesised Q&A) survives across sessions,
+hosts, and harnesses. It holds 100+ hard-won entries: infra incident postmortems,
+Go pitfalls, framework gotchas, design principles, ADRs. **It is not optional
+reference material — query it actively, not just when explicitly told.**
 
-- **MCP**: `mcp://hyperguild.<TAILNET>.ts.net:3100/knowledge`
-- **HTTP**: `http://hyperguild.<TAILNET>.ts.net:3100/api/v1/search`
+### When to query (treat as a reflex)
 
-<!-- TODO: replace <TAILNET> placeholder with the real Tailscale tailnet
-     name once hyperguild is deployed. Until then, agents that try to
-     reach the knowledge service on a host where it isn't running will
-     get DNS NXDOMAIN, which is the desired fail-loudly behavior. -->
-- **Scoping**: defaults to `public` collection; client projects filter to `{client}` + `public`
+- **Before** starting a non-trivial task — search for prior art with the symptom
+  AND the system component ("how did we solve X in Y?"). 5 seconds beats 5 hours.
+- **When debugging** — search for the error string, the stack frame, the affected
+  service. Past you may have already paid this tax.
+- **Before adopting** a pattern, library, framework, or model name — check if it
+  was tried and rejected, or what the integration footguns are.
+- **When making architectural decisions** — search for the domain + "ADR" or
+  "decision" to find prior reasoning before re-deriving it.
+- **When a recommendation feels novel** — challenge yourself: "has this been
+  documented?" The brain often has it.
+
+### When to write
+
+After you discover something that **future-you would forget** and that **isn't
+recoverable from the code, git log, or PR description alone**:
+
+- Bugs whose root cause is non-obvious and generalisable beyond this project.
+- Framework / library / model-name quirks that bit you and would bite anyone.
+- Design principles validated under fire (e.g. "every `_get` needs a `_list`").
+- Postmortems for incidents: what broke, why, how diagnosed, what to do next time.
+
+DON'T write project status, sprint progress, PR summaries, or "what I did this
+session" — those rot fast and the originals are in git/gitea anyway. Brain
+entries that age well are about *why*, *how to avoid*, and *what to do when*.
+
+### How to access (per harness)
+
+| Harness | Query | Write |
+|---------|-------|-------|
+| **Claude Code, Claude Desktop** | `brain_query` (BM25), `brain_answer` (LLM-synth + sources) MCP tools | `brain_write` MCP tool |
+| **Crush, Pi, Antigravity, other MCP-capable** | same MCP server: `ingestion-brain` (via the `mcp__*_brain__*` namespace once authenticated) | same |
+| **Anything HTTP-only (curl, scripts)** | `POST https://brain-mcp.d-ma.be/query` with `{"query":"..."}` (auth via `BRAIN_MCP_TOKEN`) | `POST .../write` with `{"content":"...","filename":"..."}` |
+| **Browser / human inspection** | `https://gitea.d-ma.be/mathias/hyperguild` → `knowledge/` and `wiki/` markdown files |
+
+- **Scoping**: defaults to `public` collection; client projects filter to `{client}` + `public`.
+- **Routing**: brain_answer's LLM uses berget.ai as primary, iguana ollama as
+  fallback. Both are configurable in the `supervisor/ingestion-deployment.yaml`
+  on the koala k3s cluster; don't hardcode local-only model names into the
+  berget URL (see knowledge entry on namespace mismatches).
+
+### Quick reflex checks
+
+If you find yourself about to say any of these out loud, you owe yourself a brain query first:
+
+- "I think the issue might be..."
+- "Let me try X and see..."
+- "I'll just write a script to..."
+- "This is probably a new bug..."
+- "Has anyone done this before?" — *yes, probably, go check.*
 
 ## Client work rules
 
@@ -208,9 +270,11 @@ Key skills:
 
 ### Git
 - Conventional commits: `feat:`, `fix:`, `chore:`, `docs:`, `refactor:`
-- Branch naming: `feat/short-description`, `fix/short-description`
-- PRs: one concern per PR, description explains *why* not *what*
-- **Branch protection:** always work on a feature branch, open a PR, never push directly to main
+- **Trunk-Based Development:** commit directly to main. One logical change per commit.
+- Run `task check` locally before every push. CI is the quality gate, not branch protection.
+- No feature branches, no PRs for solo/agent work.
+- Exception: if a parallel agent session is active on this repo, use a short-lived
+  `agent/<description>` branch and merge within the same session.
 
 ### Security
 - No secrets in code, ever — use env vars or SOPS-encrypted files
@@ -248,98 +312,29 @@ When acting as a coding agent on this project:
 3. If unsure about a convention, check `DECISIONS.md` or ask
 4. Never modify files outside the project root without explicit permission
 5. When adding a dependency, explain why in the commit message
-6. Always work on a feature branch and open a PR — never push directly to main
+6. Commit directly to main. Run `task check` before every push. Never create
+   feature branches unless a parallel agent is simultaneously active on this repo.
 7. For client projects: never send code or context to cloud APIs — use local models via LiteLLM
 
-## Current sprint — gitea-mcp v0.2 patch (2026-05-14)
+## Current state — v0.2.5 (2026-05-17)
 
-### Context
-The main v0.2 batch (repo_create, repo_update, repo_mirror_push, repo_delete,
-repo_tree, repo_topics_update, file_read dir-fix, issue_get, release_create,
-create_project_from_template) was implemented and pushed directly to main.
+All v0.2 work is complete and deployed. No active sprint.
 
-This sprint fixes three remaining gaps found during code review on 2026-05-14.
-These are blockers for `hyperguild new-project`.
+### What shipped
 
-### Issues to fix (all three in one PR: `fix/v02-patch`)
+| Tag | PR | Tools / fixes |
+|-----|----|---------------|
+| v0.2.2 | #21 | repo_create, repo_update, repo_mirror_push |
+| v0.2.3 | #22 | repo_tree, repo_topics_update, file_read dir fix |
+| v0.2.4 | #23 | issue_get, release_create, repo_delete |
+| v0.2.5 | #26 | repo_update archived+template, create_project_from_template template_name, pr_files_diff loop fix |
 
-#### #12 — repo_update: add `archived` and `template` fields
-**File:** `internal/gitea/repos.go` → `UpdateRepoArgs` struct
-**File:** `internal/tools/repo_update.go` → input schema + args struct
+Current main: `e31fd3f`. CI green. Deployed via Flux.
 
-Add to `UpdateRepoArgs`:
-```go
-Archived *bool
-Template *bool
-```
+### Next up
 
-Add to tool input schema:
-```json
-"archived": {
-  "type": "boolean",
-  "description": "Mark repo as archived (read-only). Requires confirm=<repo name>."
-},
-"template": {
-  "type": "boolean",
-  "description": "Toggle template repo flag."
-}
-```
+1. **`hyperguild new-project` v1** — primary next target.
+   See brain node `adr-new-project-gitea-first-github-mirror` for full flow spec.
 
-Add confirm-guard for `archived=true` (same pattern as `private=false`):
-```go
-if args.Archived != nil && *args.Archived {
-    if args.Confirm != args.Name {
-        return nil, fmt.Errorf("setting archived=true is irreversible: set confirm=%q to proceed", args.Name)
-    }
-}
-```
-
-New test cases to add in `repo_update_test.go`:
-- `TestRepoUpdateTool_Archive` — happy path with confirm
-- `TestRepoUpdateTool_ArchiveRequiresConfirm` — missing confirm returns error
-- `TestRepoUpdateTool_SetTemplate` — no confirm needed
-
-#### #24 — create_project_from_template: make template selectable
-**File:** `internal/tools/create_project_from_template.go`
-
-Add optional `template_name` param to input schema:
-```json
-"template_name": {
-  "type": "string",
-  "enum": ["template-go-web", "template-go-agent"],
-  "description": "Template repo to generate from. Defaults to template-go-web.",
-  "default": "template-go-web"
-}
-```
-
-The tool should use `args.TemplateName` if set, fall back to the hardcoded default.
-Remove the hardcoded template name from `cmd/gitea-mcp/main.go` constructor call —
-the tool resolves it internally.
-
-New test case: `TestCreateProjectFromTemplate_AgentTemplate`
-
-#### #25 — pr_files_diff: fix same diff returned for all files
-**File:** `internal/tools/pr_files_diff.go`
-
-There is a loop bug where all file entries in the response contain the same diff
-(the first file's diff is reused for every subsequent file). Find the loop and
-ensure each iteration reads and assigns the correct diff for its own file.
-
-Reproduce: call `pr_files_diff` on any PR with 3+ files, verify each file has
-a distinct diff.
-
-### Definition of done
-
-- [ ] `task check` passes
-- [ ] `repo_update` accepts `archived` and `template` params
-- [ ] `archived=true` requires `confirm=<repo name>`
-- [ ] `create_project_from_template` accepts `template_name` param, defaults to `template-go-web`
-- [ ] `pr_files_diff` returns distinct diff per file
-- [ ] All new test cases pass
-- [ ] PR `fix/v02-patch` merged to main via PR (not direct push)
-
-### After this sprint
-
-Next: `hyperguild new-project` v1 implementation.
-See brain node `adr-new-project-gitea-first-github-mirror` for the full flow spec.
-Also: verify end-to-end mirror flow (issue #19) once `repo_mirror_push` is confirmed working.
+2. **Issue #19** — end-to-end mirror flow verification.
+   `repo_mirror_push` is implemented but the full flow (create repo → add push mirror → verify sync to GitHub) has not been tested manually. Do this before relying on it in production.

CLAUDE.md

@@ -37,9 +37,11 @@
 
 ### Git
 - Conventional commits: `feat:`, `fix:`, `chore:`, `docs:`, `refactor:`
-- Branch naming: `feat/short-description`, `fix/short-description`
-- PRs: one concern per PR, description explains *why* not *what*
-- **Branch protection:** always work on a feature branch, open a PR, never push directly to main
+- **Trunk-Based Development:** commit directly to main. One logical change per commit.
+- Run `task check` locally before every push. CI is the quality gate, not branch protection.
+- No feature branches, no PRs for solo/agent work.
+- Exception: if a parallel agent session is active on this repo, use a short-lived
+  `agent/<description>` branch and merge within the same session.
 
 ### Security
 - No secrets in code, ever — use env vars or SOPS-encrypted files
@@ -77,98 +79,29 @@ When acting as a coding agent on this project:
 3. If unsure about a convention, check `DECISIONS.md` or ask
 4. Never modify files outside the project root without explicit permission
 5. When adding a dependency, explain why in the commit message
-6. Always work on a feature branch and open a PR — never push directly to main
+6. Commit directly to main. Run `task check` before every push. Never create
+   feature branches unless a parallel agent is simultaneously active on this repo.
 7. For client projects: never send code or context to cloud APIs — use local models via LiteLLM
 
-## Current sprint — gitea-mcp v0.2 patch (2026-05-14)
+## Current state — v0.2.5 (2026-05-17)
 
-### Context
-The main v0.2 batch (repo_create, repo_update, repo_mirror_push, repo_delete,
-repo_tree, repo_topics_update, file_read dir-fix, issue_get, release_create,
-create_project_from_template) was implemented and pushed directly to main.
+All v0.2 work is complete and deployed. No active sprint.
 
-This sprint fixes three remaining gaps found during code review on 2026-05-14.
-These are blockers for `hyperguild new-project`.
+### What shipped
 
-### Issues to fix (all three in one PR: `fix/v02-patch`)
+| Tag | PR | Tools / fixes |
+|-----|----|---------------|
+| v0.2.2 | #21 | repo_create, repo_update, repo_mirror_push |
+| v0.2.3 | #22 | repo_tree, repo_topics_update, file_read dir fix |
+| v0.2.4 | #23 | issue_get, release_create, repo_delete |
+| v0.2.5 | #26 | repo_update archived+template, create_project_from_template template_name, pr_files_diff loop fix |
 
-#### #12 — repo_update: add `archived` and `template` fields
-**File:** `internal/gitea/repos.go` → `UpdateRepoArgs` struct
-**File:** `internal/tools/repo_update.go` → input schema + args struct
+Current main: `e31fd3f`. CI green. Deployed via Flux.
 
-Add to `UpdateRepoArgs`:
-```go
-Archived *bool
-Template *bool
-```
+### Next up
 
-Add to tool input schema:
-```json
-"archived": {
-  "type": "boolean",
-  "description": "Mark repo as archived (read-only). Requires confirm=<repo name>."
-},
-"template": {
-  "type": "boolean",
-  "description": "Toggle template repo flag."
-}
-```
+1. **`hyperguild new-project` v1** — primary next target.
+   See brain node `adr-new-project-gitea-first-github-mirror` for full flow spec.
 
-Add confirm-guard for `archived=true` (same pattern as `private=false`):
-```go
-if args.Archived != nil && *args.Archived {
-    if args.Confirm != args.Name {
-        return nil, fmt.Errorf("setting archived=true is irreversible: set confirm=%q to proceed", args.Name)
-    }
-}
-```
-
-New test cases to add in `repo_update_test.go`:
-- `TestRepoUpdateTool_Archive` — happy path with confirm
-- `TestRepoUpdateTool_ArchiveRequiresConfirm` — missing confirm returns error
-- `TestRepoUpdateTool_SetTemplate` — no confirm needed
-
-#### #24 — create_project_from_template: make template selectable
-**File:** `internal/tools/create_project_from_template.go`
-
-Add optional `template_name` param to input schema:
-```json
-"template_name": {
-  "type": "string",
-  "enum": ["template-go-web", "template-go-agent"],
-  "description": "Template repo to generate from. Defaults to template-go-web.",
-  "default": "template-go-web"
-}
-```
-
-The tool should use `args.TemplateName` if set, fall back to the hardcoded default.
-Remove the hardcoded template name from `cmd/gitea-mcp/main.go` constructor call —
-the tool resolves it internally.
-
-New test case: `TestCreateProjectFromTemplate_AgentTemplate`
-
-#### #25 — pr_files_diff: fix same diff returned for all files
-**File:** `internal/tools/pr_files_diff.go`
-
-There is a loop bug where all file entries in the response contain the same diff
-(the first file's diff is reused for every subsequent file). Find the loop and
-ensure each iteration reads and assigns the correct diff for its own file.
-
-Reproduce: call `pr_files_diff` on any PR with 3+ files, verify each file has
-a distinct diff.
-
-### Definition of done
-
-- [ ] `task check` passes
-- [ ] `repo_update` accepts `archived` and `template` params
-- [ ] `archived=true` requires `confirm=<repo name>`
-- [ ] `create_project_from_template` accepts `template_name` param, defaults to `template-go-web`
-- [ ] `pr_files_diff` returns distinct diff per file
-- [ ] All new test cases pass
-- [ ] PR `fix/v02-patch` merged to main via PR (not direct push)
-
-### After this sprint
-
-Next: `hyperguild new-project` v1 implementation.
-See brain node `adr-new-project-gitea-first-github-mirror` for the full flow spec.
-Also: verify end-to-end mirror flow (issue #19) once `repo_mirror_push` is confirmed working.
+2. **Issue #19** — end-to-end mirror flow verification.
+   `repo_mirror_push` is implemented but the full flow (create repo → add push mirror → verify sync to GitHub) has not been tested manually. Do this before relying on it in production.

Dockerfile

@@ -1,5 +1,16 @@
 FROM golang:1.26-alpine AS build
 WORKDIR /src
+
+# Fetch internal gitea-hosted Go modules (e.g. mcp-chassis) without going
+# through proxy.golang.org and without HTTP→HTTPS surprises. Gitea returns
+# http:// in its go-import meta tag, so rewrite to https here and bypass
+# the module proxy + sumdb.
+RUN apk add --no-cache git && \
+    git config --global url."https://gitea.d-ma.be/".insteadOf "http://gitea.d-ma.be/"
+ENV GOPRIVATE=gitea.d-ma.be
+ENV GOPROXY=direct
+ENV GOSUMDB=off
+
 COPY go.mod go.sum ./
 RUN go mod download
 COPY . .

README.md

@@ -2,3 +2,14 @@
 
 Streamable HTTP MCP service exposing Gitea repo operations to Claude apps.
 See `~/dev/AI/infra/docs/superpowers/specs/2026-05-04-gitea-mcp-gitops-workflow-design.md`.
+
+## Quickstart
+
+```bash
+task setup:hooks   # installs .githooks/pre-push — runs task check before every push
+task check         # context sync + lint + test + vet
+task build         # produces bin/gitea-mcp
+```
+
+This repo uses Trunk-Based Development. Commit directly to `main`. The pre-push
+hook enforces the quality gate locally; CI re-runs `task check` on every push.

Taskfile.yml

@@ -47,6 +47,13 @@ tasks:
     cmds:
       - bash scripts/context-sync.sh
 
+  setup:hooks:
+    desc: Install git hooks (.githooks/pre-push)
+    cmds:
+      - git config core.hooksPath .githooks
+      - chmod +x .githooks/pre-push
+      - echo "✓ git hooks installed (pre-push runs task check)"
+
   context:sync:claude:
     cmds: [bash scripts/context-sync.sh claude]
   context:sync:agents:

cmd/gitea-mcp/main.go

@@ -2,10 +2,12 @@ package main
 
 import (
 	"context"
-	"encoding/json"
 	"log/slog"
 	"net/http"
 	"os"
+	"strings"
+
+	chassisauth "gitea.d-ma.be/mathias/mcp-chassis/auth"
 
 	"gitea.d-ma.be/mathias/gitea-mcp/internal/allowlist"
 	"gitea.d-ma.be/mathias/gitea-mcp/internal/auth"
@@ -27,7 +29,7 @@ func main() {
 
 	ctx := context.Background()
 
-	jwtValidator, err := auth.NewJWTValidator(ctx, cfg.DexIssuerURL, cfg.MCPAudience)
+	jwtValidator, err := chassisauth.NewJWTValidator(ctx, cfg.DexIssuerURL, cfg.MCPAudience)
 	if err != nil {
 		logger.Warn("jwt validator init failed; JWT auth disabled", "err", err)
 	}
@@ -66,6 +68,10 @@ func main() {
 	reg.Register(tools.NewRepoTree(giteaClient, ownerAllow))
 	reg.Register(tools.NewRepoTopicsUpdate(giteaClient, ownerAllow))
 	reg.Register(tools.NewIssueGet(giteaClient, ownerAllow))
+	reg.Register(tools.NewIssueList(giteaClient, ownerAllow))
+	reg.Register(tools.NewIssueClose(giteaClient, ownerAllow))
+	reg.Register(tools.NewIssueReopen(giteaClient, ownerAllow))
+	reg.Register(tools.NewWorkflowRunList(giteaClient, ownerAllow))
 	reg.Register(tools.NewReleaseCreate(giteaClient, ownerAllow))
 	reg.Register(tools.NewRepoDelete(giteaClient, ownerAllow))
 
@@ -74,9 +80,17 @@ func main() {
 		Sessions: mcp.NewSessionStore(),
 	})
 
+	// resourceMetadataURL is only emitted in the WWW-Authenticate challenge
+	// when both MCPResourceURL and a Dex issuer are wired; empty disables
+	// the challenge so static-only clients aren't pushed into OAuth discovery.
+	var resourceMetadataURL string
+	if cfg.MCPResourceURL != "" && cfg.DexIssuerURL != "" {
+		resourceMetadataURL = strings.TrimRight(cfg.MCPResourceURL, "/") + "/.well-known/oauth-protected-resource"
+	}
+
 	mux := http.NewServeMux()
 	mux.Handle("/mcp", mcp.OriginAllowlist(cfg.OriginAllowlist)(
-		auth.BearerMiddleware(jwtValidator, cfg.StaticToken,
+		chassisauth.BearerMiddleware(cfg.StaticToken, jwtValidator, "gitea", resourceMetadataURL,
 			auth.CallerMiddleware(mcpSrv),
 		),
 	))
@@ -84,21 +98,10 @@ func main() {
 		w.WriteHeader(http.StatusOK)
 		_, _ = w.Write([]byte("ok"))
 	})
-	mux.HandleFunc("/.well-known/oauth-protected-resource", func(w http.ResponseWriter, r *http.Request) {
-		if r.Method != http.MethodGet {
-			http.Error(w, "method not allowed", http.StatusMethodNotAllowed)
-			return
-		}
-		w.Header().Set("Content-Type", "application/json")
-		payload := map[string]any{
-			"resource":             cfg.MCPResourceURL,
-			"authorization_servers": []string{},
-		}
 	if cfg.DexIssuerURL != "" {
-			payload["authorization_servers"] = []string{cfg.DexIssuerURL}
+		mux.HandleFunc("GET /.well-known/oauth-protected-resource",
+			chassisauth.ProtectedResourceHandler(cfg.MCPResourceURL, cfg.DexIssuerURL))
 	}
-		_ = json.NewEncoder(w).Encode(payload)
-	})
 
 	addr := ":" + cfg.Port
 	logger.Info("gitea-mcp starting", "addr", addr, "version", "0.1.0")

go.mod

@@ -9,6 +9,7 @@ require (
 )
 
 require (
+	gitea.d-ma.be/mathias/mcp-chassis v0.1.0 // indirect
 	github.com/davecgh/go-spew v1.1.1 // indirect
 	github.com/decred/dcrd/dcrec/secp256k1/v4 v4.4.0 // indirect
 	github.com/goccy/go-json v0.10.3 // indirect

go.sum

@@ -1,3 +1,5 @@
+gitea.d-ma.be/mathias/mcp-chassis v0.1.0 h1:8RXO34+n7Vu8HnUMagars6fc4oemqRpMu7MVtjaj4qY=
+gitea.d-ma.be/mathias/mcp-chassis v0.1.0/go.mod h1:ajbLlwr2L7FAN3TBU39KucZkKJM02wTbKbDKDEW2YvE=
 github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=

internal/auth/bearer.go

@@ -1,42 +0,0 @@
-package auth
-
-import (
-	"crypto/subtle"
-	"net/http"
-	"strings"
-)
-
-// BearerMiddleware authenticates requests via the Authorization header.
-//
-// A request is allowed when:
-//
-//  1. The Bearer token is a valid JWT issued by the configured Dex OIDC server, or
-//  2. The Bearer token matches staticToken (constant-time compare).
-//
-// Any other case — including missing or empty Authorization header — returns 401.
-//
-// The Gitea service PAT is intentionally NOT used to authenticate the caller:
-// it is only used by the Gitea client for upstream API calls. Decoupling the
-// two prevents the MCP endpoint from being reachable anonymously when a service
-// PAT happens to be configured.
-func BearerMiddleware(jwtValidator *JWTValidator, staticToken string, next http.Handler) http.Handler {
-	return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		bearer, hasBearer := strings.CutPrefix(r.Header.Get("Authorization"), "Bearer ")
-		if !hasBearer || bearer == "" {
-			http.Error(w, "unauthorized", http.StatusUnauthorized)
-			return
-		}
-
-		if jwtValidator.Validate(r.Context(), bearer) {
-			next.ServeHTTP(w, r)
-			return
-		}
-
-		if staticToken != "" && subtle.ConstantTimeCompare([]byte(bearer), []byte(staticToken)) == 1 {
-			next.ServeHTTP(w, r)
-			return
-		}
-
-		http.Error(w, "unauthorized", http.StatusUnauthorized)
-	})
-}

internal/auth/bearer_test.go

@@ -1,92 +0,0 @@
-package auth_test
-
-import (
-	"net/http"
-	"net/http/httptest"
-	"testing"
-
-	"gitea.d-ma.be/mathias/gitea-mcp/internal/auth"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func okHandler(called *bool) http.Handler {
-	return http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
-		if called != nil {
-			*called = true
-		}
-		w.WriteHeader(http.StatusOK)
-	})
-}
-
-func TestBearerMiddleware_NoAuthHeader(t *testing.T) {
-	srv := httptest.NewServer(auth.BearerMiddleware(nil, "", okHandler(nil)))
-	defer srv.Close()
-
-	resp, err := http.Post(srv.URL+"/mcp", "application/json", nil)
-	require.NoError(t, err)
-	defer func() { _ = resp.Body.Close() }()
-	assert.Equal(t, http.StatusUnauthorized, resp.StatusCode)
-}
-
-func TestBearerMiddleware_NoAuthHeader_RejectsEvenWhenStaticConfigured(t *testing.T) {
-	// A configured staticToken must not allow unauthenticated callers through.
-	srv := httptest.NewServer(auth.BearerMiddleware(nil, "any-static", okHandler(nil)))
-	defer srv.Close()
-
-	resp, err := http.Post(srv.URL+"/mcp", "application/json", nil)
-	require.NoError(t, err)
-	defer func() { _ = resp.Body.Close() }()
-	assert.Equal(t, http.StatusUnauthorized, resp.StatusCode)
-}
-
-func TestBearerMiddleware_EmptyBearer(t *testing.T) {
-	srv := httptest.NewServer(auth.BearerMiddleware(nil, "static", okHandler(nil)))
-	defer srv.Close()
-
-	req, _ := http.NewRequest(http.MethodPost, srv.URL+"/mcp", nil)
-	req.Header.Set("Authorization", "Bearer ")
-	resp, err := http.DefaultClient.Do(req)
-	require.NoError(t, err)
-	defer func() { _ = resp.Body.Close() }()
-	assert.Equal(t, http.StatusUnauthorized, resp.StatusCode)
-}
-
-func TestBearerMiddleware_StaticToken_Valid(t *testing.T) {
-	const staticToken = "my-static-token"
-	called := false
-	srv := httptest.NewServer(auth.BearerMiddleware(nil, staticToken, okHandler(&called)))
-	defer srv.Close()
-
-	req, _ := http.NewRequest(http.MethodPost, srv.URL+"/mcp", nil)
-	req.Header.Set("Authorization", "Bearer "+staticToken)
-	resp, err := http.DefaultClient.Do(req)
-	require.NoError(t, err)
-	defer func() { _ = resp.Body.Close() }()
-	assert.Equal(t, http.StatusOK, resp.StatusCode)
-	assert.True(t, called)
-}
-
-func TestBearerMiddleware_StaticToken_Invalid(t *testing.T) {
-	srv := httptest.NewServer(auth.BearerMiddleware(nil, "correct-token", okHandler(nil)))
-	defer srv.Close()
-
-	req, _ := http.NewRequest(http.MethodPost, srv.URL+"/mcp", nil)
-	req.Header.Set("Authorization", "Bearer wrong-token")
-	resp, err := http.DefaultClient.Do(req)
-	require.NoError(t, err)
-	defer func() { _ = resp.Body.Close() }()
-	assert.Equal(t, http.StatusUnauthorized, resp.StatusCode)
-}
-
-func TestBearerMiddleware_UnknownBearer_NoStatic_NoJWT(t *testing.T) {
-	srv := httptest.NewServer(auth.BearerMiddleware(nil, "", okHandler(nil)))
-	defer srv.Close()
-
-	req, _ := http.NewRequest(http.MethodPost, srv.URL+"/mcp", nil)
-	req.Header.Set("Authorization", "Bearer random-unknown-token")
-	resp, err := http.DefaultClient.Do(req)
-	require.NoError(t, err)
-	defer func() { _ = resp.Body.Close() }()
-	assert.Equal(t, http.StatusUnauthorized, resp.StatusCode)
-}

internal/auth/jwt.go

@@ -1,79 +0,0 @@
-package auth
-
-import (
-	"context"
-	"encoding/json"
-	"fmt"
-	"net/http"
-	"time"
-
-	"github.com/lestrrat-go/jwx/v2/jwk"
-	"github.com/lestrrat-go/jwx/v2/jwt"
-)
-
-// JWTValidator validates bearer tokens as JWTs issued by a Dex OIDC server.
-// A nil JWTValidator always returns false — JWT validation is disabled.
-type JWTValidator struct {
-	issuer  string
-	aud     string
-	cache   *jwk.Cache
-	jwksURI string
-}
-
-// NewJWTValidator creates a validator by fetching the OIDC discovery document
-// from issuerURL. Returns nil, nil when issuerURL is empty (disabled).
-func NewJWTValidator(ctx context.Context, issuerURL, audience string) (*JWTValidator, error) {
-	if issuerURL == "" {
-		return nil, nil
-	}
-
-	resp, err := http.Get(issuerURL + "/.well-known/openid-configuration")
-	if err != nil {
-		return nil, fmt.Errorf("fetch oidc discovery: %w", err)
-	}
-	defer func() { _ = resp.Body.Close() }()
-
-	var doc struct {
-		JWKSURI string `json:"jwks_uri"`
-	}
-	if err := json.NewDecoder(resp.Body).Decode(&doc); err != nil {
-		return nil, fmt.Errorf("decode oidc discovery: %w", err)
-	}
-
-	cache := jwk.NewCache(ctx)
-	if err := cache.Register(doc.JWKSURI, jwk.WithRefreshInterval(time.Hour)); err != nil {
-		return nil, fmt.Errorf("register jwks uri: %w", err)
-	}
-	// warm the cache immediately so first request doesn't block
-	if _, err := cache.Refresh(ctx, doc.JWKSURI); err != nil {
-		return nil, fmt.Errorf("warm jwks cache: %w", err)
-	}
-
-	return &JWTValidator{
-		issuer:  issuerURL,
-		aud:     audience,
-		cache:   cache,
-		jwksURI: doc.JWKSURI,
-	}, nil
-}
-
-// Validate returns true if rawToken is a valid JWT signed by the OIDC server.
-func (v *JWTValidator) Validate(ctx context.Context, rawToken string) bool {
-	if v == nil {
-		return false
-	}
-	keySet, err := v.cache.Get(ctx, v.jwksURI)
-	if err != nil {
-		return false
-	}
-	opts := []jwt.ParseOption{
-		jwt.WithKeySet(keySet),
-		jwt.WithIssuer(v.issuer),
-		jwt.WithValidate(true),
-	}
-	if v.aud != "" {
-		opts = append(opts, jwt.WithAudience(v.aud))
-	}
-	_, err = jwt.Parse([]byte(rawToken), opts...)
-	return err == nil
-}

internal/gitea/issues.go

@@ -4,6 +4,8 @@ import (
 	"context"
 	"encoding/json"
 	"fmt"
+	"net/url"
+	"strconv"
 )
 
 type Issue struct {
@@ -72,6 +74,72 @@ func (c *Client) CreateIssue(ctx context.Context, owner, repo string, args Creat
 	return &iss, nil
 }
 
+// ListIssuesArgs captures the optional query params for ListIssues.
+type ListIssuesArgs struct {
+	State  string // "open" | "closed" | "all"
+	Labels string // comma-separated label names
+	Since  string // ISO 8601
+	Page   int
+	Limit  int
+}
+
+// ListIssues fetches issues for a repo. Pulls are excluded server-side
+// (type=issues) so they don't leak through the same endpoint.
+func (c *Client) ListIssues(ctx context.Context, owner, repo string, args ListIssuesArgs) ([]Issue, error) {
+	q := url.Values{}
+	q.Set("type", "issues")
+	if args.State != "" {
+		q.Set("state", args.State)
+	}
+	if args.Labels != "" {
+		q.Set("labels", args.Labels)
+	}
+	if args.Since != "" {
+		q.Set("since", args.Since)
+	}
+	if args.Page > 0 {
+		q.Set("page", strconv.Itoa(args.Page))
+	}
+	if args.Limit > 0 {
+		q.Set("limit", strconv.Itoa(args.Limit))
+	}
+	p := fmt.Sprintf("/api/v1/repos/%s/%s/issues?%s", owner, repo, q.Encode())
+	body, status, err := c.GetJSON(ctx, p)
+	if err != nil {
+		return nil, err
+	}
+	if err := MapStatus(status, body); err != nil {
+		return nil, err
+	}
+	var issues []Issue
+	if err := json.Unmarshal(body, &issues); err != nil {
+		return nil, err
+	}
+	return issues, nil
+}
+
+// SetIssueState flips an issue between "open" and "closed" via PATCH.
+// Gitea uses the same endpoint for both transitions.
+func (c *Client) SetIssueState(ctx context.Context, owner, repo string, number int, state string) (*Issue, error) {
+	p := fmt.Sprintf("/api/v1/repos/%s/%s/issues/%d", owner, repo, number)
+	payload, err := json.Marshal(map[string]string{"state": state})
+	if err != nil {
+		return nil, err
+	}
+	body, status, err := c.PatchJSON(ctx, p, payload)
+	if err != nil {
+		return nil, err
+	}
+	if err := MapStatus(status, body); err != nil {
+		return nil, err
+	}
+	var iss Issue
+	if err := json.Unmarshal(body, &iss); err != nil {
+		return nil, err
+	}
+	return &iss, nil
+}
+
 type IssueComment struct {
 	ID      int64  `json:"id"`
 	Body    string `json:"body"`

internal/gitea/workflows.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"encoding/json"
 	"fmt"
+	"net/url"
 	"strconv"
 	"strings"
 )
@@ -55,10 +56,20 @@ func (c *Client) DispatchWorkflow(ctx context.Context, owner, repo, workflow str
 // WorkflowRun represents a Gitea Actions run.
 type WorkflowRun struct {
 	ID           int64  `json:"id"`
+	DisplayTitle string `json:"display_title,omitempty"`
 	Status       string `json:"status"`     // queued | in_progress | completed
 	Conclusion   string `json:"conclusion"` // success | failure | cancelled | skipped (only when completed)
+	Event        string `json:"event,omitempty"`
+	HeadSHA      string `json:"head_sha,omitempty"`
+	HeadBranch   string `json:"head_branch,omitempty"`
+	WorkflowID   string `json:"workflow_id,omitempty"`
+	RunNumber    int64  `json:"run_number,omitempty"`
 	StartedAt    string `json:"started_at"`
+	UpdatedAt    string `json:"updated_at,omitempty"`
 	HTMLURL      string `json:"html_url"`
+	Actor        struct {
+		Login string `json:"login"`
+	} `json:"actor,omitempty"`
 }
 
 // GetWorkflowRun fetches the status of a specific Actions run.
@@ -77,3 +88,59 @@ func (c *Client) GetWorkflowRun(ctx context.Context, owner, repo string, runID i
 	}
 	return &run, nil
 }
+
+// ListWorkflowRunsArgs captures the optional query params for ListWorkflowRuns.
+type ListWorkflowRunsArgs struct {
+	Branch   string
+	HeadSHA  string
+	Status   string // queued | in_progress | completed | all
+	Event    string // push | pull_request | schedule | workflow_dispatch | all
+	Workflow string
+	Page     int
+	Limit    int
+}
+
+type workflowRunsResponse struct {
+	TotalCount   int64         `json:"total_count"`
+	WorkflowRuns []WorkflowRun `json:"workflow_runs"`
+}
+
+// ListWorkflowRuns fetches recent Actions runs for a repo with optional filters.
+// Status / Event of "all" or "" are treated as no-filter.
+func (c *Client) ListWorkflowRuns(ctx context.Context, owner, repo string, args ListWorkflowRunsArgs) (*workflowRunsResponse, error) {
+	q := url.Values{}
+	if args.Branch != "" {
+		q.Set("branch", args.Branch)
+	}
+	if args.HeadSHA != "" {
+		q.Set("head_sha", args.HeadSHA)
+	}
+	if args.Status != "" && args.Status != "all" {
+		q.Set("status", args.Status)
+	}
+	if args.Event != "" && args.Event != "all" {
+		q.Set("event", args.Event)
+	}
+	if args.Workflow != "" {
+		q.Set("workflow", args.Workflow)
+	}
+	if args.Page > 0 {
+		q.Set("page", strconv.Itoa(args.Page))
+	}
+	if args.Limit > 0 {
+		q.Set("limit", strconv.Itoa(args.Limit))
+	}
+	p := fmt.Sprintf("/api/v1/repos/%s/%s/actions/runs?%s", owner, repo, q.Encode())
+	body, status, err := c.GetJSON(ctx, p)
+	if err != nil {
+		return nil, err
+	}
+	if err := MapStatus(status, body); err != nil {
+		return nil, err
+	}
+	var resp workflowRunsResponse
+	if err := json.Unmarshal(body, &resp); err != nil {
+		return nil, err
+	}
+	return &resp, nil
+}

internal/tools/issue_close.go

@@ -0,0 +1,56 @@
+package tools
+
+import (
+	"context"
+	"encoding/json"
+
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/allowlist"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/gitea"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/registry"
+)
+
+type IssueClose struct {
+	c *gitea.Client
+	a *allowlist.Allowlist
+}
+
+func NewIssueClose(c *gitea.Client, a *allowlist.Allowlist) *IssueClose {
+	return &IssueClose{c: c, a: a}
+}
+
+func (t *IssueClose) Descriptor() registry.ToolDescriptor {
+	return registry.ToolDescriptor{
+		Name:        "issue_close",
+		Description: "Close an open issue. Reversible via issue_reopen.",
+		InputSchema: json.RawMessage(`{
+			"type":"object",
+			"properties":{
+				"owner":{"type":"string"},
+				"name":{"type":"string"},
+				"number":{"type":"integer","minimum":1}
+			},
+			"required":["owner","name","number"]
+		}`),
+	}
+}
+
+type issueCloseArgs struct {
+	Owner  string `json:"owner"`
+	Name   string `json:"name"`
+	Number int    `json:"number"`
+}
+
+func (t *IssueClose) Call(ctx context.Context, raw json.RawMessage) (json.RawMessage, error) {
+	var args issueCloseArgs
+	if err := parseArgs(raw, &args); err != nil {
+		return nil, err
+	}
+	if err := t.a.Check(args.Owner); err != nil {
+		return nil, err
+	}
+	iss, err := t.c.SetIssueState(ctx, args.Owner, args.Name, args.Number, "closed")
+	if err != nil {
+		return nil, err
+	}
+	return textOK(iss)
+}

internal/tools/issue_close_test.go

@@ -0,0 +1,52 @@
+package tools_test
+
+import (
+	"context"
+	"encoding/json"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/allowlist"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/gitea"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/tools"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestIssueCloseTool(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		assert.Equal(t, http.MethodPatch, r.Method)
+		assert.Equal(t, "/api/v1/repos/mathias/infra/issues/26", r.URL.Path)
+		b, _ := io.ReadAll(r.Body)
+		assert.JSONEq(t, `{"state":"closed"}`, string(b))
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{"number":26,"title":"feat: ntfy via NPM","state":"closed","html_url":"http://gitea.example.com/mathias/infra/issues/26"}`))
+	}))
+	defer srv.Close()
+
+	tool := tools.NewIssueClose(gitea.NewClient(srv.URL, "tok"), allowlist.New([]string{"mathias"}))
+	out, err := tool.Call(context.Background(), json.RawMessage(`{"owner":"mathias","name":"infra","number":26}`))
+	require.NoError(t, err)
+	assert.Contains(t, string(out), `"number":26`)
+	assert.Contains(t, string(out), `"state":"closed"`)
+}
+
+func TestIssueCloseTool_NotFound(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(http.StatusNotFound)
+		_, _ = w.Write([]byte(`{"message":"issue not found"}`))
+	}))
+	defer srv.Close()
+
+	tool := tools.NewIssueClose(gitea.NewClient(srv.URL, "tok"), allowlist.New([]string{"mathias"}))
+	_, err := tool.Call(context.Background(), json.RawMessage(`{"owner":"mathias","name":"infra","number":999}`))
+	require.Error(t, err)
+}
+
+func TestIssueCloseAllowlistRejects(t *testing.T) {
+	tool := tools.NewIssueClose(gitea.NewClient("http://unused", ""), allowlist.New([]string{"mathias"}))
+	_, err := tool.Call(context.Background(), json.RawMessage(`{"owner":"evil","name":"x","number":1}`))
+	require.Error(t, err)
+}

internal/tools/issue_list.go

@@ -0,0 +1,83 @@
+package tools
+
+import (
+	"context"
+	"encoding/json"
+
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/allowlist"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/gitea"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/registry"
+)
+
+type IssueList struct {
+	c *gitea.Client
+	a *allowlist.Allowlist
+}
+
+func NewIssueList(c *gitea.Client, a *allowlist.Allowlist) *IssueList {
+	return &IssueList{c: c, a: a}
+}
+
+func (t *IssueList) Descriptor() registry.ToolDescriptor {
+	return registry.ToolDescriptor{
+		Name:        "issue_list",
+		Description: "List issues in a repo with optional filters. PRs are excluded (use pr_list for those).",
+		InputSchema: json.RawMessage(`{
+			"type":"object",
+			"properties":{
+				"owner":{"type":"string"},
+				"name":{"type":"string"},
+				"state":{"type":"string","enum":["open","closed","all"]},
+				"labels":{"type":"string"},
+				"since":{"type":"string"},
+				"page":{"type":"integer","minimum":1},
+				"limit":{"type":"integer","minimum":1,"maximum":50}
+			},
+			"required":["owner","name"]
+		}`),
+	}
+}
+
+type issueListArgs struct {
+	Owner  string `json:"owner"`
+	Name   string `json:"name"`
+	State  string `json:"state"`
+	Labels string `json:"labels"`
+	Since  string `json:"since"`
+	Page   int    `json:"page"`
+	Limit  int    `json:"limit"`
+}
+
+func (t *IssueList) Call(ctx context.Context, raw json.RawMessage) (json.RawMessage, error) {
+	var args issueListArgs
+	if err := parseArgs(raw, &args); err != nil {
+		return nil, err
+	}
+	if err := t.a.Check(args.Owner); err != nil {
+		return nil, err
+	}
+	if args.State == "" {
+		args.State = "open"
+	}
+	args.Limit = capLimit(args.Limit, 30)
+	if args.Page < 1 {
+		args.Page = 1
+	}
+	issues, err := t.c.ListIssues(ctx, args.Owner, args.Name, gitea.ListIssuesArgs{
+		State:  args.State,
+		Labels: args.Labels,
+		Since:  args.Since,
+		Page:   args.Page,
+		Limit:  args.Limit,
+	})
+	if err != nil {
+		return nil, err
+	}
+	out := map[string]any{
+		"issues": issues,
+	}
+	if len(issues) == args.Limit {
+		out["next_page"] = args.Page + 1
+	}
+	return textOK(out)
+}

internal/tools/issue_list_test.go

@@ -0,0 +1,88 @@
+package tools_test
+
+import (
+	"context"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/allowlist"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/gitea"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/tools"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestIssueListTool(t *testing.T) {
+	tests := []struct {
+		name      string
+		input     string
+		wantQuery map[string]string
+		respBody  string
+		assert    func(t *testing.T, out string)
+	}{
+		{
+			name:      "happy path defaults",
+			input:     `{"owner":"mathias","name":"infra"}`,
+			wantQuery: map[string]string{"type": "issues", "state": "open", "page": "1", "limit": "30"},
+			respBody:  `[{"number":42,"title":"fix auth","state":"open","html_url":"http://gitea.example/m/infra/issues/42"},{"number":41,"title":"add tests","state":"open"}]`,
+			assert: func(t *testing.T, out string) {
+				assert.Contains(t, out, `"number":42`)
+				assert.Contains(t, out, `"number":41`)
+			},
+		},
+		{
+			name:      "state filter",
+			input:     `{"owner":"mathias","name":"infra","state":"closed"}`,
+			wantQuery: map[string]string{"type": "issues", "state": "closed"},
+			respBody:  `[]`,
+			assert: func(t *testing.T, out string) {
+				assert.Contains(t, out, `"issues":[]`)
+			},
+		},
+		{
+			name:      "label + since filter",
+			input:     `{"owner":"mathias","name":"infra","labels":"bug,critical","since":"2026-05-01T00:00:00Z"}`,
+			wantQuery: map[string]string{"labels": "bug,critical", "since": "2026-05-01T00:00:00Z"},
+			respBody:  `[]`,
+			assert:    func(t *testing.T, out string) {},
+		},
+		{
+			name:      "empty result",
+			input:     `{"owner":"mathias","name":"infra"}`,
+			wantQuery: map[string]string{"state": "open"},
+			respBody:  `[]`,
+			assert: func(t *testing.T, out string) {
+				assert.Contains(t, out, `"issues":[]`)
+				assert.NotContains(t, out, `next_page`)
+			},
+		},
+	}
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+				assert.Equal(t, http.MethodGet, r.Method)
+				assert.Equal(t, "/api/v1/repos/mathias/infra/issues", r.URL.Path)
+				q := r.URL.Query()
+				for k, v := range tc.wantQuery {
+					assert.Equal(t, v, q.Get(k), "query param %q", k)
+				}
+				w.Header().Set("Content-Type", "application/json")
+				_, _ = w.Write([]byte(tc.respBody))
+			}))
+			defer srv.Close()
+
+			tool := tools.NewIssueList(gitea.NewClient(srv.URL, "tok"), allowlist.New([]string{"mathias"}))
+			out, err := tool.Call(context.Background(), json.RawMessage(tc.input))
+			require.NoError(t, err)
+			tc.assert(t, string(out))
+		})
+	}
+}
+
+func TestIssueListAllowlistRejects(t *testing.T) {
+	tool := tools.NewIssueList(gitea.NewClient("http://unused", ""), allowlist.New([]string{"mathias"}))
+	_, err := tool.Call(context.Background(), json.RawMessage(`{"owner":"evil","name":"x"}`))
+	require.Error(t, err)
+}

internal/tools/issue_reopen.go

@@ -0,0 +1,56 @@
+package tools
+
+import (
+	"context"
+	"encoding/json"
+
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/allowlist"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/gitea"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/registry"
+)
+
+type IssueReopen struct {
+	c *gitea.Client
+	a *allowlist.Allowlist
+}
+
+func NewIssueReopen(c *gitea.Client, a *allowlist.Allowlist) *IssueReopen {
+	return &IssueReopen{c: c, a: a}
+}
+
+func (t *IssueReopen) Descriptor() registry.ToolDescriptor {
+	return registry.ToolDescriptor{
+		Name:        "issue_reopen",
+		Description: "Reopen a closed issue. Reversible via issue_close.",
+		InputSchema: json.RawMessage(`{
+			"type":"object",
+			"properties":{
+				"owner":{"type":"string"},
+				"name":{"type":"string"},
+				"number":{"type":"integer","minimum":1}
+			},
+			"required":["owner","name","number"]
+		}`),
+	}
+}
+
+type issueReopenArgs struct {
+	Owner  string `json:"owner"`
+	Name   string `json:"name"`
+	Number int    `json:"number"`
+}
+
+func (t *IssueReopen) Call(ctx context.Context, raw json.RawMessage) (json.RawMessage, error) {
+	var args issueReopenArgs
+	if err := parseArgs(raw, &args); err != nil {
+		return nil, err
+	}
+	if err := t.a.Check(args.Owner); err != nil {
+		return nil, err
+	}
+	iss, err := t.c.SetIssueState(ctx, args.Owner, args.Name, args.Number, "open")
+	if err != nil {
+		return nil, err
+	}
+	return textOK(iss)
+}

internal/tools/issue_reopen_test.go

@@ -0,0 +1,40 @@
+package tools_test
+
+import (
+	"context"
+	"encoding/json"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/allowlist"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/gitea"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/tools"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestIssueReopenTool(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		assert.Equal(t, http.MethodPatch, r.Method)
+		assert.Equal(t, "/api/v1/repos/mathias/infra/issues/26", r.URL.Path)
+		b, _ := io.ReadAll(r.Body)
+		assert.JSONEq(t, `{"state":"open"}`, string(b))
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{"number":26,"title":"feat: ntfy via NPM","state":"open","html_url":"http://gitea.example.com/mathias/infra/issues/26"}`))
+	}))
+	defer srv.Close()
+
+	tool := tools.NewIssueReopen(gitea.NewClient(srv.URL, "tok"), allowlist.New([]string{"mathias"}))
+	out, err := tool.Call(context.Background(), json.RawMessage(`{"owner":"mathias","name":"infra","number":26}`))
+	require.NoError(t, err)
+	assert.Contains(t, string(out), `"number":26`)
+	assert.Contains(t, string(out), `"state":"open"`)
+}
+
+func TestIssueReopenAllowlistRejects(t *testing.T) {
+	tool := tools.NewIssueReopen(gitea.NewClient("http://unused", ""), allowlist.New([]string{"mathias"}))
+	_, err := tool.Call(context.Background(), json.RawMessage(`{"owner":"evil","name":"x","number":1}`))
+	require.Error(t, err)
+}

internal/tools/workflow_run_list.go

@@ -0,0 +1,87 @@
+package tools
+
+import (
+	"context"
+	"encoding/json"
+
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/allowlist"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/gitea"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/registry"
+)
+
+type WorkflowRunList struct {
+	c *gitea.Client
+	a *allowlist.Allowlist
+}
+
+func NewWorkflowRunList(c *gitea.Client, a *allowlist.Allowlist) *WorkflowRunList {
+	return &WorkflowRunList{c: c, a: a}
+}
+
+func (t *WorkflowRunList) Descriptor() registry.ToolDescriptor {
+	return registry.ToolDescriptor{
+		Name:        "workflow_run_list",
+		Description: "List recent Gitea Actions workflow runs with optional filters (branch, head_sha, status, event, workflow).",
+		InputSchema: json.RawMessage(`{
+			"type":"object",
+			"properties":{
+				"owner":{"type":"string"},
+				"name":{"type":"string"},
+				"branch":{"type":"string"},
+				"head_sha":{"type":"string"},
+				"status":{"type":"string","enum":["queued","in_progress","completed","all"]},
+				"event":{"type":"string","enum":["push","pull_request","schedule","workflow_dispatch","all"]},
+				"workflow":{"type":"string"},
+				"page":{"type":"integer","minimum":1},
+				"limit":{"type":"integer","minimum":1,"maximum":50}
+			},
+			"required":["owner","name"]
+		}`),
+	}
+}
+
+type workflowRunListArgs struct {
+	Owner    string `json:"owner"`
+	Name     string `json:"name"`
+	Branch   string `json:"branch"`
+	HeadSHA  string `json:"head_sha"`
+	Status   string `json:"status"`
+	Event    string `json:"event"`
+	Workflow string `json:"workflow"`
+	Page     int    `json:"page"`
+	Limit    int    `json:"limit"`
+}
+
+func (t *WorkflowRunList) Call(ctx context.Context, raw json.RawMessage) (json.RawMessage, error) {
+	var args workflowRunListArgs
+	if err := parseArgs(raw, &args); err != nil {
+		return nil, err
+	}
+	if err := t.a.Check(args.Owner); err != nil {
+		return nil, err
+	}
+	args.Limit = capLimit(args.Limit, 10)
+	if args.Page < 1 {
+		args.Page = 1
+	}
+	resp, err := t.c.ListWorkflowRuns(ctx, args.Owner, args.Name, gitea.ListWorkflowRunsArgs{
+		Branch:   args.Branch,
+		HeadSHA:  args.HeadSHA,
+		Status:   args.Status,
+		Event:    args.Event,
+		Workflow: args.Workflow,
+		Page:     args.Page,
+		Limit:    args.Limit,
+	})
+	if err != nil {
+		return nil, err
+	}
+	out := map[string]any{
+		"runs":  resp.WorkflowRuns,
+		"total": resp.TotalCount,
+	}
+	if len(resp.WorkflowRuns) == args.Limit {
+		out["next_page"] = args.Page + 1
+	}
+	return textOK(out)
+}

internal/tools/workflow_run_list_test.go

@@ -0,0 +1,98 @@
+package tools_test
+
+import (
+	"context"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/allowlist"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/gitea"
+	"gitea.d-ma.be/mathias/gitea-mcp/internal/tools"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestWorkflowRunListTool(t *testing.T) {
+	tests := []struct {
+		name      string
+		input     string
+		wantQuery map[string]string
+		notQuery  []string
+		respBody  string
+		assert    func(t *testing.T, out string)
+	}{
+		{
+			name:      "happy path defaults",
+			input:     `{"owner":"mathias","name":"gitea-mcp"}`,
+			wantQuery: map[string]string{"page": "1", "limit": "10"},
+			respBody:  `{"total_count":2,"workflow_runs":[{"id":823,"status":"completed","conclusion":"success","head_sha":"dc907fb"},{"id":822,"status":"completed","conclusion":"success","head_sha":"c4bd339"}]}`,
+			assert: func(t *testing.T, out string) {
+				assert.Contains(t, out, `"id":823`)
+				assert.Contains(t, out, `"total":2`)
+			},
+		},
+		{
+			name:      "head_sha short filter",
+			input:     `{"owner":"mathias","name":"gitea-mcp","head_sha":"dc907fb"}`,
+			wantQuery: map[string]string{"head_sha": "dc907fb"},
+			respBody:  `{"total_count":1,"workflow_runs":[{"id":823,"status":"completed","conclusion":"success","head_sha":"dc907fb"}]}`,
+			assert: func(t *testing.T, out string) {
+				assert.Contains(t, out, `"id":823`)
+			},
+		},
+		{
+			name:      "status filter",
+			input:     `{"owner":"mathias","name":"gitea-mcp","status":"in_progress"}`,
+			wantQuery: map[string]string{"status": "in_progress"},
+			respBody:  `{"total_count":0,"workflow_runs":[]}`,
+			assert: func(t *testing.T, out string) {
+				assert.Contains(t, out, `"runs":[]`)
+			},
+		},
+		{
+			name:      "status=all is no-op",
+			input:     `{"owner":"mathias","name":"gitea-mcp","status":"all"}`,
+			notQuery:  []string{"status"},
+			respBody:  `{"total_count":0,"workflow_runs":[]}`,
+			assert:    func(t *testing.T, out string) {},
+		},
+		{
+			name:      "branch filter",
+			input:     `{"owner":"mathias","name":"gitea-mcp","branch":"main"}`,
+			wantQuery: map[string]string{"branch": "main"},
+			respBody:  `{"total_count":0,"workflow_runs":[]}`,
+			assert:    func(t *testing.T, out string) {},
+		},
+	}
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+				assert.Equal(t, http.MethodGet, r.Method)
+				assert.Equal(t, "/api/v1/repos/mathias/gitea-mcp/actions/runs", r.URL.Path)
+				q := r.URL.Query()
+				for k, v := range tc.wantQuery {
+					assert.Equal(t, v, q.Get(k), "query param %q", k)
+				}
+				for _, k := range tc.notQuery {
+					assert.Equal(t, "", q.Get(k), "query param %q should be absent", k)
+				}
+				w.Header().Set("Content-Type", "application/json")
+				_, _ = w.Write([]byte(tc.respBody))
+			}))
+			defer srv.Close()
+
+			tool := tools.NewWorkflowRunList(gitea.NewClient(srv.URL, "tok"), allowlist.New([]string{"mathias"}))
+			out, err := tool.Call(context.Background(), json.RawMessage(tc.input))
+			require.NoError(t, err)
+			tc.assert(t, string(out))
+		})
+	}
+}
+
+func TestWorkflowRunListAllowlistRejects(t *testing.T) {
+	tool := tools.NewWorkflowRunList(gitea.NewClient("http://unused", ""), allowlist.New([]string{"mathias"}))
+	_, err := tool.Call(context.Background(), json.RawMessage(`{"owner":"evil","name":"x"}`))
+	require.Error(t, err)
+}