marketing-plan
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an inherent attack surface for indirect prompt injection as it is designed to ingest and summarize large volumes of external, potentially untrusted data to generate a marketing plan.
- Ingestion points: The skill processes user-provided materials in the
materials/directory and pulls live data from multiple external API integrations including Stripe, Ahrefs, GA4, and Customer.io (referenced inreferences/methodology.md). - Capability inventory: The skill has the capability to write aggregated content to the local filesystem (
~/marketing-plans/) and can optionally publish the final plan to external GitHub repositories via git commit and push operations. - Boundary markers: The instructions do not define explicit boundary markers or delimiters for external content to prevent the model from following instructions that might be embedded within the gathered research data.
- Sanitization: While the skill includes a 'verification pass' (Phase 3) to ensure brand voice consistency and remove machine-specific paths, it lacks explicit logic for sanitizing or filtering untrusted input for malicious prompt injection patterns.
Audit Metadata