prepare-regression
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is instructed to ingest untrusted data from external sources, specifically PR titles, bodies, discussions, reviews, and inline comments.
- Ingestion points: External Pull Request metadata and discussion content (SKILL.md, Step 2).
- Boundary markers: No specific delimiters or instructions to ignore embedded commands are provided to separate the external PR content from the agent's instructions.
- Capability inventory: The skill executes a local script
../../scripts/snapshot.pyand writes to a localprepare.jsonfile (SKILL.md, Step 3 & 4). - Sanitization: No explicit sanitization, filtering, or validation of the ingested PR content is described.
- [COMMAND_EXECUTION]: The skill executes a local Python script using the
uvtool. - Evidence:
uv run --no-project ../../scripts/snapshot.pyis called with arguments including the repository and commit identifiers derived from user input (SKILL.md, Step 3).
Audit Metadata