code-review
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's logic relies on reading and following instructions found in external, potentially untrusted files to define its goals and acceptance criteria.
- Ingestion points: The skill reads
workflow/<change-slug>/SPEC.md,PLAN.md, andREVIEW.md, and analyzesgit diffoutputs. Malicious instructions could be embedded in these files (e.g., as comments or requirements). - Boundary markers: There are no defined delimiters or 'ignore embedded instructions' warnings provided to distinguish between the skill's operational logic and the data it processes.
- Capability inventory: The skill is authorized to perform file writes (
REVIEW.md) and execute shell commands for project verification. - Sanitization: No sanitization or validation of the content within the
SPEC.mdor code changes is performed before the agent treats it as authoritative for the review process. - [COMMAND_EXECUTION]: The skill directs the agent to run local shell commands (
test,lint,typecheck,build) to verify code. While standard for development, this execution surface can be exploited if the underlying project configuration or test scripts are modified by an attacker.
Audit Metadata