cawplan-product-report
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill demonstrates high security awareness by incorporating a dedicated reference file (references/UNTRUSTED_CONTENT.md) that explicitly warns against following instructions found within ticket data. It enforces strict title fidelity while treating the content as non-executable data.
- [COMMAND_EXECUTION]: The skill uses the Bash tool to interact with the cawplan CLI. It includes significant constraints on command construction (no shell syntax, no pipes, no quotes) within references/QUERY_RESILIENCE.md to prevent command injection and ensure predictable behavior in hosted environments.
- [INDIRECT_PROMPT_INJECTION]: While the skill processes user-generated ticket content, it effectively mitigates injection risks through specific instruction sets. Ingestion points: Ticket metadata and descriptions are ingested via cawplan tickets search and cawplan tickets history in various workflows (A, B, C, D). Boundary markers: The references/UNTRUSTED_CONTENT.md file serves as a comprehensive boundary marker, defining what content is untrusted. Capability inventory: The skill's capabilities are limited to reading data through the cawplan CLI; no automated write operations are performed without user confirmation. Sanitization: The skill instructs the agent to quote content rather than merging it and to explicitly ignore any instructions found in the data.
Audit Metadata