writing-plans
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted specifications to generate implementation plans containing executable shell commands and code. Malicious instructions within the source specifications could potentially be propagated into the final plan.\n
- Ingestion points: User-provided requirements or specifications (processed in SKILL.md).\n
- Boundary markers: None specified for the input data; the skill relies on the agent's contextual understanding of the 'spec'.\n
- Capability inventory: The skill writes plan documents to the local file system. These documents contain tasks meant for execution by other high-capability skills (e.g., executing-plans).\n
- Sanitization: No explicit sanitization or filtering of input specifications is implemented.
Audit Metadata