comet-review
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes project documentation, specifications, and untracked code files from the local workspace, creating an entry point for potential indirect prompt injections.
- Ingestion points: Reads
proposal.md,design.md,tasks.md,specs/*/spec.md, and arbitrary untracked files in the workspace (documented inSKILL.mdSections 2 and 3). - Boundary markers: Absent; there are no specific markers or instructions directing the agent to disregard potential instructions embedded within the codebase or documentation under review.
- Capability inventory: Invokes local tools via
comet status,comet state get,comet native show, andgit status. The skill explicitly restricts writing or executing arbitrary project code. - Sanitization: No sanitization, filtering, or escaping is performed on the files read from the codebase.
Audit Metadata