audit-stale-comments

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it analyzes untrusted content (source code comments) which could contain malicious instructions designed to influence the agent's behavior.
  • Ingestion points: JavaScript, TypeScript, and Go source code comments (SKILL.md).
  • Boundary markers: Absent; no delimiters are defined to separate comment content from analysis instructions.
  • Capability inventory: File discovery (git ls-files, rg --files), file reading, and subagent delegation. It is limited to a read-only context and explicitly forbids file edits without separate user permission (SKILL.md).
  • Sanitization: Absent; the agent evaluates raw comment text for behavioral claims.
  • [COMMAND_EXECUTION]: The skill directs the agent to use git ls-files and rg --files for repository-aware file discovery. These tools are used appropriately for their intended purpose within the current working directory.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 12:28 PM
Security Audit — agent-trust-hub — audit-stale-comments