suede-marketing-plan
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [SAFE]: The skill operates within a controlled three-phase workflow (INIT, REVIEW, FINALIZE) that maintains high levels of user oversight. It explicitly prohibits the invention of data and requires authorization for any external mutation or publication.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external APIs (Stripe, Ahrefs, Google Analytics) and user-provided documents in the
materials/directory. This creates a surface area for indirect prompt injection. However, the skill mitigates this by enforcing an interactive 'REVIEW' phase where the user must approve every section of the plan before it is compiled, and a 'Verification Pass' (Step 3.2) that checks for accuracy and unintended content. - [EXTERNAL_DOWNLOADS]: The skill references integrations with well-known services including Stripe, Shopify, GitHub, and Google Analytics to fetch business metrics. These references are neutral and standard for the skill's purpose, with instructions explicitly directing the agent to use only pre-authorized, secure connections.
- [DATA_EXPOSURE]: While the skill accesses sensitive business information (revenue, customer LTV, marketing budgets), it does so using the agent's standardized Model Context Protocol (MCP) tools. It includes clear boundaries against hardcoding credentials or accessing unauthorized system paths.
Audit Metadata