autoreview
Pass
Audited by Gen Agent Trust Hub on Aug 17, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes the 'autoreview' and 'gh' CLI tools. It mitigates shell injection risks by instructing the agent to pipe multiline task contracts via standard input (--prompt-file -) rather than passing untrusted data as command-line arguments.- [DATA_EXFILTRATION]: Comprehensive instructions in references/security.md ensure that any data shared externally (such as for bug reporting) is sanitized. The guidelines explicitly forbid the inclusion of private source code, credentials, environment variables, or raw provider outputs.- [PROMPT_INJECTION]: The skill handles untrusted external data (PRs, issues, and code) by requiring the agent to distill authoritative criteria into a controlled task contract. It specifically warns against executing embedded instructions found in linked repository content.- [EXTERNAL_DOWNLOADS]: The skill requires 'autoreview' and 'trufflehog' binaries but explicitly prohibits the agent from downloading or installing them itself, directing the user to use trusted host package managers or internal release workflows.
Audit Metadata