requesting-code-review
Pass
Audited by Gen Agent Trust Hub on Mar 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell commands like
git rev-parse,git log, andgit diffwith variable interpolation inSKILL.mdandcode-reviewer.md. If git metadata (e.g., branch names) or user-provided placeholders contain shell metacharacters, it could lead to unintended command execution. - [PROMPT_INJECTION]: The
code-reviewer.mdtemplate is vulnerable to indirect prompt injection because it interpolates external inputs such as{DESCRIPTION}and{PLAN_REFERENCE}directly into the instructions. An attacker could provide malicious content to influence or bypass the reviewer agent's logic. - Ingestion points: Placeholders in
code-reviewer.md({WHAT_WAS_IMPLEMENTED},{PLAN_OR_REQUIREMENTS},{DESCRIPTION},{PLAN_REFERENCE}). - Boundary markers: None identified; untrusted content is mixed directly with system-level instructions.
- Capability inventory: The skill uses subprocess calls to
gitfor repository analysis. - Sanitization: No input sanitization or escaping is performed on the interpolated strings.
Audit Metadata