Refactor AGENTS.md: Extract detailed instructions to prevent context pollution

Fixes #340

Created AGENT_INSTRUCTIONS.md to house detailed operational instructions,
reducing AGENTS.md from 963 to 626 lines (~35% reduction).

Changes:
- New AGENT_INSTRUCTIONS.md with development, testing, and release procedures
- AGENTS.md now serves as quick overview with references to detailed docs
- Updated README.md link to "Landing the Plane" section
- Fixed broken team-workflow README links

Benefits:
- Prevents context pollution when multiple tools add instructions
- Better separation: quick reference vs detailed operations
- Enables tool-specific instruction files
- Maintains all content, just reorganized

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
Steve Yegge
2025-11-20 22:04:17 -05:00
parent be37a13ceb
commit 21a0656a4c
4 changed files with 412 additions and 366 deletions

View File

@@ -245,7 +245,7 @@ This pattern has proven invaluable for maintaining database hygiene and preventi
**5. Choose next work**
- Provide a formatted prompt for the next session with context
See the ["Landing the Plane"](AGENTS.md#landing-the-plane) section in this project's `AGENTS.md` for a complete example you can adapt. The key insight: explicitly reminding agents to maintain issue tracker hygiene prevents the common problem of agents creating issues during work but forgetting to sync them at session end.
See the ["Landing the Plane"](AGENT_INSTRUCTIONS.md#landing-the-plane) section in this project's documentation for a complete example you can adapt. The key insight: explicitly reminding agents to maintain issue tracker hygiene prevents the common problem of agents creating issues during work but forgetting to sync them at session end.
## The Magic: Distributed Database via Git