Thanks for helping improve FableCodex. This project is a Codex-native workflow plugin, so contributions should preserve its main boundary: it can improve Codex operating habits, but it must not claim to unlock or reproduce the Fable 5 model.
- Improve documentation clarity or examples.
- Add tests for scripts, manifests, and reference files.
- Tighten provider-bridge guidance when official provider behavior changes.
- Add small, verifiable workflow helpers.
- Improve coverage accounting without copying large upstream prompt passages.
This repository intentionally uses the Python standard library for tests.
python -m unittest discover -s tests -v
python plugins/codex-fable5/skills/codex-fable5/scripts/fable_coverage.pyBefore opening a pull request, also run:
python -m py_compile \
plugins/codex-fable5/skills/codex-fable5/scripts/codex_goals.py \
plugins/codex-fable5/skills/codex-fable5/scripts/fable_coverage.py \
plugins/codex-fable5/skills/codex-fable5/scripts/make_litellm_config.py \
tests/test_scripts.py- Keep changes scoped and evidence-backed.
- Do not paste long source prompt passages or proprietary model instructions.
- Do not claim Claude, Anthropic, Fable, or Mythos identity unless an actual configured provider proves that context.
- Keep provider setup optional and credential-free.
- Do not commit
.codex-fable5/local task ledgers or secrets. - Prefer stdlib-only tests unless a dependency materially reduces risk.
- The change matches the project boundary and does not overpromise model capability.
- Tests pass locally.
- Any current product, model, API, or provider claim is verified against an official source.
- Documentation is updated when user-facing behavior changes.
- Licensing and attribution notes remain accurate.