content-blog-post
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data from business profiles, content kits, and customer-provided "proof banks" to generate blog content. This ingestion of external data creates a potential surface for indirect prompt injection.
- Ingestion points: External data enters the context via
lib/business.ts,lib/business-profile.ts, and referenced "proof bank" files. - Boundary markers: The skill uses structured Markdown tables to define brief fields but does not implement explicit LLM boundary markers (e.g., XML tags) for interpolated data.
- Capability inventory: The skill possesses the ability to write, list, delete, and restore posts via the
write_post,list_posts,delete_post, andrestore_posttools. - Sanitization: The risk is mitigated by a mandatory human-in-the-loop requirement ("Never publish without the owner's approval") and a secondary quality check process (
content-publish-check) before final approval. - [DATA_EXPOSURE]: The skill explicitly instructs the agent to "Strip GPS location from photo files before upload," which is a proactive measure to prevent the accidental exposure of sensitive location data.
- [SAFE]: The skill includes clear "Stop rules" to prevent the generation of content without verifiable first-hand proof, ensuring content integrity and preventing the generation of low-quality or deceptive material.
Audit Metadata