Conventions for human contributors and AI agents working on this repository.
## Style
- Terse technical prose. No emojis in commits, issues, PR comments, or code.
- Python: PEP 484 type hints. Hook script stays in standard-library only — no pip dependencies.
- Tabs for indentation in JSON and Markdown tables. Spaces (4) for Python.
- Double quotes for JSON strings.
## Layout
-`hooks/ultrawork-detector.py` — pure stdlib `UserPromptSubmit` hook. Reads JSON on stdin, writes the directive to stdout when the keyword matches, exits 0 otherwise.
-`agents/*.toml` — bundled Codex agent role files. Installed into `CODEX_HOME/agents/` by `src/cli/install-codex/link-cached-plugin-agents.ts` at install time (symlink on Unix, copy on Windows). No runtime `SessionStart` hook is involved.
-`hooks/hooks.json` — registers the prompt-detector hook only.
-`.codex-plugin/plugin.json` — Codex plugin manifest. Marketplace metadata lives here, not in `package.json`.
## Constraints
- Never let the hook block a turn — exit code is always 0.
- Never make a network call from the hook.
- Keep the directive in `ULTRAWORK_DIRECTIVE` self-contained inside the Python file. The prompt hook is a single artifact a reviewer can read top-to-bottom.
- Keep bundled agent role prompts concise and model-specific; measure prompt length when changing them.
- When editing `ULTRAWORK_DIRECTIVE`, apply the `prompt-engineering` skill's entropy gate: every edit must reduce uncertainty per token. Re-measure character count before committing.
