code-reviewer
Pass
Audited by Gen Agent Trust Hub on May 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to run tools found within the environment of the project being reviewed, specifically 'the project's configured duplicate-code check'. Executing commands based on the configuration of untrusted code repositories presents a risk where a malicious project could define a 'duplicate-code' command that executes harmful payloads.
- [PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection (Category 8). It ingests untrusted data from source code files and git diffs and uses this data to drive its analysis and verification workflow. An attacker could embed malicious instructions within code comments or documentation to influence the agent's behavior.
- Ingestion points: Workflow Step 2 (READ) and Step 3 (CONTEXT) ingest target files and git diffs.
- Boundary markers: Absent. The instructions do not specify delimiters or warnings to ignore instructions found within the analyzed code.
- Capability inventory: The agent has file-read access and the capability to execute shell commands (
rg,bugmagnet, and project-specific duplicate checks). - Sanitization: Absent. The skill does not provide any logic for filtering or escaping content found in the untrusted files before processing.
Audit Metadata