staged-task-execution
Pass
Audited by Gen Agent Trust Hub on Apr 8, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a structured task management system designed with human-in-the-loop oversight as a core requirement.
- [COMMAND_EXECUTION]: Task execution is performed one atomic step at a time and is gated by a mandatory confirmation step using the vscode_askQuestions tool, which prevents unauthorized or bulk execution of commands.
- [DATA_EXFILTRATION]: The skill manages session state using a local temp/ directory to track progress and context; it does not access sensitive user credentials or system files, and contains no network exfiltration logic.
- [PROMPT_INJECTION]: The skill processes user-provided reference materials. Ingestion points: temp/references/ folder. Boundary markers: Absent. Capability inventory: File writing and code implementation. Sanitization: Absent. Risk is mitigated by the Step 7 mandatory confirmation for every task.
Audit Metadata