appsec
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted repositories, PR diffs, and configuration files. Maliciously crafted content within these inputs could attempt to override the agent's instructions or bias its security analysis.
- Ingestion points: Filesystem access to project source code and configuration files (.env, .yml, .json) during the
Reviewphase defined inSKILL.md. - Boundary markers: The instructions mandate verifying all external claims against the actual source code and utilizing a formal rubric and scoring formula to ensure findings are cited against established security principles (A1-A43).
- Capability inventory: The skill performs read-only filesystem operations and executes local security CLI tools. It allows for dynamic testing only on user-specified, non-production environments.
- Sanitization: There is a strict requirement to redact all sensitive values (credentials, keys, tokens) before they are included in any output or report, as specified in the
R1inventory section. - [COMMAND_EXECUTION]: The skill utilizes several standard security tools for static analysis (e.g., Semgrep, Bandit, Gosec, Gitleaks, TruffleHog). These are executed via the command line to scan project files, which is consistent with the primary purpose of the skill.
Audit Metadata