requesting-code-review

Pass

Audited by Gen Agent Trust Hub on Aug 11, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted external data (source code changes and plan requirements) through a subagent, creating a surface for indirect prompt injection. Malicious instructions embedded in code comments or documentation could potentially influence the reviewer's assessment.\n
  • Ingestion points: Data from the git repository (via git diff) and the [PLAN_OR_REQUIREMENTS] variable in SKILL.md and code-reviewer.md.\n
  • Boundary markers: The subagent prompt uses Markdown headers (e.g., ## Requirements / Plan) to delineate input data from instructions.\n
  • Capability inventory: The subagent is instructed to use git utilities including diff, show, log, and worktree.\n
  • Sanitization: The skill does not implement specific sanitization or escaping for the interpolated placeholders.\n- [COMMAND_EXECUTION]: The skill executes various git commands to manage SHAs and inspect repository history. These operations are restricted to the local repository and the subagent is explicitly instructed to operate in a read-only manner, mitigating the risk of accidental or malicious modification of the working tree.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 11, 2026, 09:44 PM
Security Audit — agent-trust-hub — requesting-code-review