skills/ascend-ai-coding/awesome-ascend-skills/external-cannbot-ops-pypto-intent-understand/Gen Agent Trust Hub
external-cannbot-ops-pypto-intent-understand
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including user-provided URLs, research papers, and code snippets, to generate specification documents (SPEC.md).
- Ingestion points: Untrusted data enters the context through natural language inputs, URL references to papers/documentation, and code fragments provided by the user in Phase 1 (Input Classification).
- Capability inventory: The skill is capable of performing network fetches (via tools like WebFetch to retrieve external materials) and writing to the local file system to generate the final SPEC.md file.
- Boundary markers: The skill incorporates a mandatory 'Visualization Confirmation' phase (Phase 2) which uses an internal 'AskUserQuestion' tool to present extracted data to the user for verification. It also explicitly labels information confidence levels (High, Medium, Low) and requires a manual 'Confirm' action before proceeding.
- Sanitization: The instructions do not specify automated sanitization, validation, or escaping of the content retrieved from external sources before it is interpreted by the agent.
Audit Metadata