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
Risk Level
SAFE
Analyzed
Jun 23, 2026, 12:46 PM
Security Audit — agent-trust-hub — task-review-workflow