squad-code-review

Pass

Audited by Gen Agent Trust Hub on Oct 23, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill contains a strong safety directive against the automatic installation of skills, plugins, MCP servers, or packages, which prevents potential supply-chain attacks or unauthorized software execution.
  • [SAFE]: It implements a zero-trust approach to data processing by explicitly instructing the agent to treat code comments, pull request text, logs, and documentation as untrusted data.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted content (diffs and PR metadata), creating an attack surface for indirect prompt injection. This risk is addressed through explicit boundary instructions and safety protocols.
  • Ingestion points: PR diffs, commit messages, code comments, logs, and external documentation.
  • Boundary markers: Present; the skill instructs the agent to handle implementation inputs as untrusted data and explicitly forbids following instructions embedded within those inputs.
  • Capability inventory: The agent is permitted to perform empirical verification but is restricted from editing features or making unauthorized mutations.
  • Sanitization: Relies on logical review gates and the comparison of findings against authoritative documentation.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 23, 2026, 06:31 PM