Add support for more AI agents in signatures.
The classic honesty trap ("I did not answer truthfully") relies on AI agents mindlessly checking every box — but modern agents are smart enough to read the checkbox and skip it. This release flips the approach: instead of tricking AI into revealing itself, rely on its honesty.
A new honest self-identification trap strategy uses a true statement like "This PR was authored and submitted by an AI agent without human review." A human won't check it (it's false for them). An honest AI agent should (it's true).
- Multi-pattern honesty trap. The
honesty-trapinput now accepts pipe-separated patterns. Repos can use multiple trap strategies simultaneously. - New default patterns. The default now ships with both the classic trick question (
I did not answer truthfully) and the new self-identification trap (This PR was authored and submitted by an AI agent without human review). - Better trap comment. When a trap fires, the comment now quotes the specific checkbox that was checked and points to the project's contributing guidelines.
- This repo's own PR template now uses the honest self-identification strategy.
Fully backwards compatible. Existing single-string honesty-trap values work without changes. To add multiple traps, separate them with |.
We recommend updating your PR template to use the new self-identification trap. Replace the old checkbox:
-- [ ] I did not answer truthfully to ALL the above checkboxes.
+- [ ] This PR was authored and submitted by an AI agent without human reviewFirst release. PR guardrails that encourage human ownership of AI-assisted contributions.
- Detects AI co-author lines in commits and AI-generated footers in PR descriptions
- Checks for unfilled PR template sections and unchecked checklist items
- Honesty trap checkbox that catches people who check every box without reading
- Welcomes first-time contributors with a pointer to contributing guidelines
- Works as a system of four files:
CONTRIBUTING.md,AGENTS.md, PR template, and this CI action