paperclip-converting-plans-to-tasks
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill consists exclusively of natural language instructions for planning and task decomposition. No executable code, shell scripts, or binary assets are provided.
- [SAFE]: No network requests, remote code patterns, or external downloads were identified. All operations are confined to the agent's internal logic and interactions with the Paperclip platform.
- [SAFE]: The skill does not access sensitive local files, credentials, or environment secrets.
- [INDIRECT_PROMPT_INJECTION]: The skill acts as a processor for user-provided planning data, which is an untrusted ingestion point. However, the instructions are limited to administrative task management within a scoped environment. Evidence: 1. Ingestion points: User-provided plans described in SKILL.md; 2. Boundary markers: Absent; 3. Capability inventory: Task creation, issue status updates, and assignment; 4. Sanitization: Not mentioned.
Audit Metadata