gs-find-bad-tests
Pass
Audited by Gen Agent Trust Hub on Nov 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it ingests and processes untrusted source code and test files from the user repository to perform analysis.
- Ingestion points: The skill reads all files within the user-defined or branch-derived scope (
SKILL.md, Step 1). - Boundary markers: The instructions do not specify any markers or delimiters to ignore potential instructions embedded within the analyzed code files.
- Capability inventory: The skill is restricted to reading files, executing informational Git commands, and outputting a report. It contains an explicit instruction to change no files.
- Sanitization: No sanitization of the ingested code content is specified before the AI processes it.
- [COMMAND_EXECUTION]: The skill uses standard, non-destructive Git commands to establish repository context and verify test history.
- Evidence: Includes commands like
git symbolic-ref --short refs/remotes/origin/HEAD,git remote show origin, andgit logto identify the default branch and check for regression tests. - [DATA_EXPOSURE]: The skill reads local source code and test suites to perform its primary function.
- Evidence: Step 1 of
SKILL.mddirects the agent to read tests and the production code they exercise. This activity is local and consistent with the stated purpose of a test auditor, with no observed exfiltration to external domains.
Audit Metadata