marketing-council
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources such as web scrapes and emails.
- Ingestion points: SKILL.md and references/skill-contract.md mention processing scraped text and inbox content.
- Boundary markers: The instructions explicitly mandate treating such data as untrusted and ignoring embedded commands.
- Capability inventory: No scripts are included in the skill; the contract requires explicit user authorization for side effects.
- Sanitization: The skill uses negative constraints to prevent the agent from following external instructions.
- [EXTERNAL_DOWNLOADS]: The skill documentation references official integrations with well-known marketing and data services.
- Evidence: references/connectors.md lists tools like Google Search Console, Ahrefs, and HubSpot.
- [SAFE]: The skill enforces strong secret management and data integrity rules.
- Evidence: Explicitly forbids hardcoding API keys in source control and requires provenance labeling for all evidence.
Audit Metadata