# Mushi Mushi — evolution-loop coding rules for Cursor
#
# These rules integrate Mushi's closed-loop bug intelligence into your
# AI-assisted development workflow. The evolution loop:
#
#   User feels a bug → Mushi captures it → AI triages it → AI opens a PR
#   → QA verifies → Judge scores → Lesson library remembers
#
# Run `mushi sync-lessons` to pull the latest project-specific rules into
# .mushi/lessons.json. The MCP server also injects them dynamically.

## Before writing a fix

1. Call the `get_fix_context` MCP tool for the report — get root cause,
   blast radius, and relevant files before touching any code.
2. Call `lessons.query` or read `.mushi/lessons.json` — apply every rule
   whose pattern matches the file area you're changing.
3. Prefer the **smallest change** that makes the test pass. Do not
   refactor unrelated code in the same commit.
4. Check `get_blast_radius` when the report touches a shared component —
   one component touching many reports means your change has higher stakes.

## After writing a fix

1. Call `submit_fix_result` with the branch, PR URL, files changed, and
   lines changed. The judge batch scores the fix overnight.
2. High-frequency lessons surface in /admin/lessons. If the same class of
   bug recurs with `iterations > 2`, the lesson is escalated to P0.

## MCP tools available (ask Cursor to call them)

| Tool                | When to use |
|---------------------|-------------|
| `get_recent_reports`| Start of a session — see what users are hitting right now |
| `get_fix_context`   | Before writing a fix — root cause + blast radius brief |
| `search_reports`    | "Are there other reports like this?" |
| `get_similar_bugs`  | Nearest-neighbour search for pattern recurrence |
| `dispatch_fix`      | Hand off a triaged report to the fix orchestrator |
| `triage_next_steps` | "What should I fix next?" — 5-item plan from the dashboard |
| `run_nl_query`      | Natural-language questions about your project data |

## Mushi lesson library

<!-- Lessons are synced from .mushi/lessons.json by `mushi sync-lessons` -->
<!-- Run that command after each deploy to keep these rules current -->