Skip to content

CLAUDE.md

Latest commit

 

History

History
45 lines (31 loc) · 2.41 KB

CLAUDE.md

File metadata and controls

45 lines (31 loc) · 2.41 KB
Raw

agy-plugin-codex — Claude Code Guide

Global-First Rule

Global instructions are authoritative:

  • ~/.claude/rules/ — behavior, routing, skill map
  • ~/.claude/skills/ — global skills

Project Context (auto-loaded every session)

  • @.ai/memory-protocol.md
  • @agent-memory/CONTEXT.md
  • @agent-memory/INDEX.md

Project Summary

Codex plugin for the Agy (Antigravity CLI) ecosystem. Mirrors Codex CLI plugin structure under ~/.codex/.

Skill Routing (domain-based, not keyword-based)

For ANY work in this project, invoke skill based on domain — do not wait for keyword:

Domain Skill
Any bug / crash / unexpected behavior debugging/debug-mantra
Plugin architecture / design decision architect
Skill file authoring or sync read rules/skills-sync-protocol.md
Any new feature / code implementation aidlc
Code review review-personas

Session Start

  1. Check agent-memory/CONTEXT.md → derive active work domain → invoke matching skill above.
  2. If task is done from previous session: update CONTEXT.md (status → idle) + append to MEMORY.md first.

graphify

This project has a knowledge graph at graphify-out/ with god nodes, community structure, and cross-file relationships.

Rules:

  • For codebase questions, first run graphify query "<question>" when graphify-out/graph.json exists. Use graphify path "<A>" "<B>" for relationships and graphify explain "<ClassName/FileName>" for a known symbol/file (name match, not free-form concept search - use query for that). These return a scoped subgraph, usually much smaller than GRAPH_REPORT.md or raw grep output.
  • If graphify-out/wiki/index.md exists, use it for broad navigation instead of raw source browsing.
  • Read graphify-out/GRAPH_REPORT.md only for broad architecture review or when query/path/explain do not surface enough context.
  • After modifying code, run graphify update . to keep the graph current (AST-only, no API cost).
  • After judging a query/path/explain result useful, a dead end, or wrong, run graphify save-result --question "Q" --answer "A" --outcome useful|dead_end|corrected --nodes N1 N2 - this accumulates across sessions so the same dead end or vocabulary mismatch isn't re-derived every time. At the start of a session, check graphify-out/reflections/LESSONS.md if it exists (built via graphify reflect) for preferred sources, known dead ends, and past corrections.