skill-reviewer
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Command Execution: The skill directs the agent to execute local shell commands via NPM, specifically
npm run tokensandnpm run test:vally. These are intended for token budget verification and integration testing within the repository environment. - Indirect Prompt Injection Surface: By design, the skill ingests and processes untrusted content from external pull request files. This is a common attack surface for AI agents, as instructions hidden within the files being reviewed could theoretically attempt to override the agent's logic.
- Ingestion points: SKILL.md, reference files, and test files identified during the 'Collect' phase of the workflow.
- Boundary markers: The instructions do not specify any delimiters or safety warnings to help the agent distinguish between its own instructions and the content of the files under review.
- Capability inventory: The skill has the ability to read arbitrary files in the repository and execute local NPM scripts.
- Sanitization: No explicit sanitization, filtering, or validation of the PR content is described prior to analysis.
Audit Metadata