instruction-docs-synclisted

# Instruction Docs Sync ## Purpose Keep project instruction files current after meaningful work without creating a second source of truth or violating the effective project policy for agent files. This skill runs after Serena memory sync and before quality checks, commits, and GitHub sync. ## Required Files - `AGENTS.md`: Codex-native project instructions. - `.claude/CLAUDE.md`: Claude Code-native project memory. These files are tracked normally on the `main` branch as ordinary source. Project policy may set `instruction_docs.mode=disabled` to skip review, but by default they are tracked project files. ## Source Of Truth Use only verified facts from: - Current code, config, scripts, manifests, hooks, workflows, tests, and git diffs. - Relevant Serena memories after they are current for the latest code. - Commands that actually exist and checks that actually run. - Official tool documentation when the file describes Codex or Claude Code behavior. Do not copy chat history, future plans, speculation, secrets, tokens, cookies, or local credentials. ## Workflow 1. Run `plugins/rldyour-flow/scripts/instruction_docs_state.py --json` or the cache copy when working outside this repository. 2. Inspect changed files and recent commits to decide whether durable project facts changed. 3. Update `AGENTS.md` for Codex: - repository purpose and source-of-truth paths; - Codex-specific plugin/skill/tool routing; - setup, validation, release, deploy, and git commands; -