Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
55 lines (42 loc) · 2.46 KB

CONTRIBUTING.md

File metadata and controls

55 lines (42 loc) · 2.46 KB
Raw

Contributing to AET

Thanks for considering a contribution. AET is intentionally small: evidence must be inspectable, local behavior must be deterministic by default, and an UNKNOWN must never be marketed as a pass.

Good ways to help

  • Report a reproducible false positive, false negative, or unsafe boundary.
  • Contribute a small positive/negative fixture for a rule or evidence contract.
  • Share an anonymized, reproducible AET workflow from a real coding-agent handoff or repository-onboarding task.
  • Improve the English or Chinese documentation, examples, or accessibility.

Please search existing Issues first. Use the Bug Report or Feature Request form so maintainers can reproduce the question without requesting sensitive repository contents.

Local development

git clone https://github.com/AdvancingTitans/agent-engineering-toolkit.git
cd agent-engineering-toolkit
uv run --no-editable --reinstall-package agent-engineering-toolkit python -m unittest discover -s tests -v
uv run --no-editable aet audit . --strict

Use Python 3.11 or later. The project intentionally uses the standard library for its runtime behavior; please discuss a new dependency before adding one.

Pull request expectations

  1. Start from a focused Issue for non-trivial changes.
  2. Keep the change narrow and explain the user-facing evidence boundary.
  3. For behavior changes, add both a passing and failing regression case.
  4. Run the commands above. For distribution changes, also run uv build and smoke-test the generated wheel in an isolated environment.
  5. Update documentation, contracts, and CHANGELOG.md when public behavior changes.

For agent-authored changes, create a focused aet.intent.json and run aet review . --base main. If you execute a proof command, use aet trace with an explicit --proof binding. Do not attach raw secrets, unredacted logs, or private repository history to an Issue or pull request.

Scope and design principles

  • Prefer a direct finding with source evidence over a broad trust score.
  • Keep audit, review, local evolve, and triage read-only.
  • Preserve explicit opt-in for generic command execution and remote retrieval.
  • Make unsupported facts UNKNOWN; do not infer private author intent.
  • Prefer a small, tested rule over a speculative framework.

By contributing, you agree that your contributions are available under this repository's MIT License.