requesting-code-review

Pass

Audited by Gen Agent Trust Hub on Oct 1, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the local repository and plan documentation, creating a surface for indirect prompt injection attacks.
  • Ingestion points: The subagent instructions in code-reviewer.md involve executing git diff and reading {PLAN_OR_REQUIREMENTS} and {DESCRIPTION} placeholders, which incorporate content from the codebase and external plan files.
  • Boundary markers: The git diff output is suggested to be viewed within a bash code block context, providing minimal structural separation that does not prevent an LLM from following instructions embedded within the diffed files.
  • Capability inventory: The subagent's capabilities are limited to reading repository state (git) and providing textual analysis. It lacks the ability to execute network requests, write to the filesystem, or access environment secrets, which significantly limits the impact of a successful injection.
  • Sanitization: No specific sanitization or filtering is applied to the output of the git commands or the interpolated description text before it is processed by the reviewer agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 1, 2026, 12:55 PM