content-strategy
Pass
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXPOSURE]: The skill attempts to read local configuration files like
.agents/product-marketing-context.md. This is a legitimate feature for retrieving business context but involves reading from the local filesystem. - [PROMPT_INJECTION]: The skill processes untrusted external data including keyword exports, customer call transcripts, and forum research from Reddit or Quora. This represents a surface for indirect prompt injection.
- Ingestion points: Accesses
.agents/product-marketing-context.md, user-provided transcripts, CSV/Excel keyword exports, and web search results. - Boundary markers: None specified to delimit or sanitize untrusted content.
- Capability inventory: No code execution or file-system writing capabilities are present in the skill instructions.
- Sanitization: No explicit sanitization or validation of the ingested data is performed.
Audit Metadata