Tool / Test impact analysis
TDAD
Builds an AST graph of Python calls, imports and test-to-code edges, then hands the agent a ranked list of tests to run for its changed files before it submits; a CLI and an agent skill.
The summary describes what the project says it does; inclusion is not an endorsement.
It moves test selection into the agent's own loop, before CI. The accompanying paper (arXiv 2603.17973) reports 70% fewer regressions on SWE-bench Verified, so CI capacity goes to PRs more likely to pass.
- Verify
AI-drafted