lean-canvas
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input (business descriptions and Lean Canvas data) without explicit sanitization or instructions to ignore embedded commands.\n
- Ingestion points: User-supplied business descriptions and canvas text within SKILL.md prompts.\n
- Boundary markers: The skill uses simple brackets for interpolation but lacks instructions for the agent to disregard instructions within the data.\n
- Capability inventory: No dangerous capabilities such as network access, file writing, or command execution are requested or present in the skill instructions.\n
- Sanitization: No input validation or filtering is specified.\n- [METADATA_POISONING]: The skill's metadata and documentation contain conflicting information.\n
- The author listed in the frontmatter (ClawFu) differs from the author context (guia-matthieu).\n
- The Skill Boundaries section incorrectly describes the skill as being for "audio production" and "audio engineering," which is irrelevant to its primary function as a business strategy tool. This indicates a failure in documentation quality or potential deceptive template reuse.
Audit Metadata