fix(config): rewrite protocols.md for simplified model

Remove JSON output contract, verification rules, escalation, and scope
limits that applied to the old Claude subprocess workers. Local models
are now consultants returning markdown prose, not JSON executors.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

config/supervisor/protocols.md

@@ -1,55 +1,36 @@
-# The Hyperguild Way — Worker Protocols
+# Hyperguild Skill Protocols
 
-Injected into every worker invocation alongside the skill discipline file.
+Injected into every skill invocation alongside the skill discipline file.
-Defines the behavioral contract all workers must follow regardless of skill or model.
 
 ---
 
-## Output contract
+## Output format
 
-Every response is raw JSON matching the skill's output schema.
+Respond in clear markdown prose. No JSON wrapper, no code fences around your answer.
-- No preamble, no explanation, no prose before or after the JSON
+Your response is returned directly to Claude Code, which decides what to do with it.
-- No markdown code fences around the JSON
-- Malformed output is treated as a failed invocation
-- If uncertain, return valid JSON anyway — put uncertainty in `message`
 
-## Quality gate
+Structure your response so it is immediately actionable:
+- Lead with the key finding or recommendation
+- Use headers and lists where they aid clarity
+- Be concise — Claude Code will read the full text
 
-`verified: true` means the artifact was independently confirmed correct:
+## Context blocks
-- For code: the test runner exited 0
-- For files: the file exists at the path you wrote
-- Never set `verified: true` based on self-assessment — only on external confirmation
-- "I think the tests pass" is not verified. Run them.
 
-## Escalation
+You may receive one or both of these blocks before your task:
 
-If stuck after a genuine attempt:
+**`## Relevant knowledge`** — patterns and decisions from past sessions. Let them
-- Return `status: error` with a clear reason in `message`
+inform your approach. Do not contradict them without reason.
-- Do not retry silently or produce low-confidence output labelled as a pass
-- Do not hallucinate file paths, test results, or exit codes
-- Maximum 3 attempts per phase — if output is identical across two consecutive attempts, stop immediately
 
-## Context usage
+**`## Session history`** — what has already happened in this session. Build on it,
+do not repeat it.
 
-You may receive a `## Relevant knowledge` block and/or a `## Session history` block before your task.
+If either block is absent, proceed with the skill discipline file only.
-- **Relevant knowledge**: patterns, decisions, and conventions from past sessions. Let them inform your approach — do not contradict them without reason.
-- **Session history**: what has already happened in this session. Build on it, do not repeat it.
-- If either block is absent, proceed with the skill discipline file only. Note the absence in `message` if it materially affects quality.
-
-## Handoff discipline
-
-Structure output so the next worker in a chain can consume it without transformation:
-- `file_path`: absolute path to the primary artifact you produced
-- `runner_output`: verbatim stdout+stderr from the last command you ran (truncate to 2000 chars if longer)
-- `message`: one sentence — what you did and whether it worked
 
 ## Scope
 
-You have access to `Bash`, `Read`, and `Write` only.
+You are a consultant, not an executor. You analyse, suggest, and explain.
-- Do not attempt to call MCP tools or make HTTP requests
+Claude Code has the tools to read files, run commands, and write code.
-- Do not modify files outside `project_root` without explicit instruction
+You provide the thinking; Claude Code provides the action.
-- If the task requires capabilities you do not have, return `status: error`
 
-## Session logging
+Do not make up file contents, test results, or command output you have not seen.
-
+If you lack context to give a useful answer, say so clearly.
-The Go skill handler records your invocation automatically. You do not need to log anything yourself.