misakanet-failure-memory.mdc

---
description: Search MisakaNet failure-memory before debugging common errors
globs:
  - "**/*.py"
  - "**/*.js"
  - "**/*.ts"
  - "**/*.yml"
  - "**/*.yaml"
  - "**/*.json"
  - "**/*.md"
---

# MisakaNet Failure Memory

When a command, test, CI, DCO, token, pip, MCP, or Windows encoding issue fails, search MisakaNet before attempting a fix from general knowledge.

## When to trigger

- Command exits with non-zero status
- Test fails with an error message
- CI/CD pipeline fails
- DCO sign-off check fails
- GitHub token / authentication errors
- pip install timeout or SSL errors
- MCP server connection issues
- Windows encoding (GBK/Unicode) errors
- Database locked / SQLite errors
- Docker build failures

## Steps

1. **Search MisakaNet** with the error message or key terms:
   ```
   misakanet_search(query="<error message or key terms>")
   ```

2. **Read the matching lesson** if results are returned:
   ```
   misakanet_get_lesson(path="<lesson path from search results>")
   ```

3. **Apply only the relevant fix** from the lesson. Do not invent fixes beyond what the lesson describes.

4. **If no match found**, submit a redacted intake:
   ```
   python3 scripts/misaka_capture.py --summary "<short error description>" --source cursor
   ```

## What NOT to do

- Do not skip MisakaNet search for known failure patterns (DCO, token, pip, MCP, encoding)
- Do not upload raw logs, prompts, or secrets to MisakaNet
- Do not auto-retry more than twice without searching MisakaNet
- Do not apply fixes from MisakaNet lessons to unrelated problems

## Example

**User:** "My pytest is failing with ImportError"

**Agent:**
1. Search: `misakanet_search(query="ImportError pytest")`
2. If lesson found → read and apply fix
3. If no lesson → `python3 scripts/misaka_capture.py --summary "pytest ImportError" --source cursor`