This repository defines an AI coding-agent skill and plugin package. Keep public guidance focused on how to use the tool itself.
- Keep
SKILL.mdandskills/agent-lean-safe-coding/SKILL.mdaligned. - Run
npm testafter changing skill text, plugin metadata, benchmark fixtures, scripts, README, or docs. - Do not commit credentials, real local paths, private logs, screenshots, generated caches, or machine-specific state.
- Use placeholders such as
<PROJECT_DIR>,<USER_HOME>, andexample.comin examples. - Keep benchmark claims tied to the reproducible fixture benchmark unless a fresh live-agent benchmark is added.
- This package may mention Codex, Claude Code, AI agents, Windows-safe editing, lean coding, and reproducible benchmark fixtures.
- Do not include private planning notes, local workspace rules, account tokens, local machine paths, or unpublished business strategy.