find-regression-commits
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Git repositories, including commit messages and diffs, and executes test oracles that could potentially be influenced by malicious content in the repository history.\n
- Ingestion points: SKILL.md (Procedure steps) and references/git-regression-search-concepts-contracts-and-invariants.md describe analyzing repository history and running test oracles.\n
- Boundary markers: The skill contains explicit instructions to use isolated worktrees or disposable clones and warns against running untrusted historical code with broad credentials or network access (in references/git-regression-search-organizational-controls-and-scale.md).\n
- Capability inventory: The skill uses scripts/bisect_oracle.py to execute external test commands via subprocess.\n
- Sanitization: The bisect_oracle.py helper script avoids shell execution and maps exit codes to Git-specific statuses to prevent infrastructure failures from being misclassified as product defects.\n- [COMMAND_EXECUTION]: The scripts/bisect_oracle.py script is designed to execute arbitrary test commands passed as arguments to serve as a test oracle for git bisect.\n
- Evidence: The script uses subprocess.run(command, ...) where command is derived from user-supplied arguments. This is the primary intended function of the tool.\n- [DYNAMIC_EXECUTION]: The integration test suite dynamically generates and executes small Python scripts to verify the behavior of the oracle helper.\n
- Evidence: scripts/test_bisect_integration.py uses sys.executable -c to run verification logic against the repository state during tests.
Audit Metadata