stabilize-tests
Pass
Audited by Gen Agent Trust Hub on Apr 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to run test suites and execute git commands (commit, push). This involves executing local code and interacting with the system, which is standard for a development-focused skill.
- [PROMPT_INJECTION]: The skill analyzes test logs and source code, creating an indirect prompt injection surface where malicious test results could attempt to influence the agent.
- Ingestion points: Test failure logs and repository source code are processed during root cause investigation.
- Boundary markers: None present to delimit untrusted test output.
- Capability inventory: Subprocess execution (running tests, git) and file system modification (applying fixes).
- Sanitization: No explicit validation or sanitization of ingested test output is performed.
Audit Metadata