task-review-workflow
Pass
Audited by Gen Agent Trust Hub on Jun 23, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill involves processing external data from pull request descriptions and linked tasks, which serves as a potential ingestion surface for indirect prompt injection. However, no malicious behavior or override instructions were found. Ingestion points: Pull request descriptions and linked task content (SKILL.md). Boundary markers: None specified. Capability inventory: Local execution of test, lint, and build commands and branch management (SKILL.md). Sanitization: No explicit sanitization or validation of the external content is mentioned.
- [COMMAND_EXECUTION]: The workflow suggests running local commands for testing and linting. These are standard development operations and do not include any suspicious, hardcoded, or arbitrarily injected malicious commands.
- [DATA_EXFILTRATION]: No evidence of unauthorized network operations, data harvesting, or credential exfiltration was identified.
Audit Metadata