marketing-loops
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONPERSISTENCE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references an external documentation file at https://github.com/coreyhaines31/marketingskills/blob/main/skills/seo-audit/references/rankings-push.md. This repository is not associated with a trusted vendor.\n- [INDIRECT_PROMPT_INJECTION]: The skill framework is designed to process untrusted data from the web, including news stories, competitor sites, and social media mentions.\n
- Ingestion points: Data enters the system via loops defined in references/loop-catalog.md (e.g., newsjacking, social-listening loops).\n
- Boundary markers: Explicit instructions are provided in references/loop-catalog.md to treat external content as untrusted and to never follow instructions found within it.\n
- Capability inventory: The agent has the capability to generate drafts for emails and social posts, and to maintain local state files.\n
- Sanitization: The skill mandates summarization and diffing of external content rather than direct execution.\n- [PERSISTENCE]: The skill utilizes a persistence mechanism to maintain state across different execution cycles.\n
- Evidence: Instructions in references/loop-state.md guide the agent to store run state in .agents/loops/ JSON and log files.\n
- Purpose: This storage tracks watermarks, handled items, and cooldowns to ensure idempotency and prevent repetitive actions between runs.
Audit Metadata