language
Pass
Audited by Gen Agent Trust Hub on Mar 11, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill code and instructions are entirely focused on marketing research and copywriting methodology. No malicious patterns, persistence mechanisms, or unauthorized data access were found.
- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it uses the WebFetch tool to read external content from sites like Reddit and Hacker News. Evidence: 1. Ingestion points: WebFetch is used to research community discussions (SKILL.md). 2. Boundary markers: The skill does not explicitly instruct the agent to use delimiters or ignore instructions within fetched text. 3. Capability inventory: The skill allows the use of Write and Edit tools, which could be misused if the agent obeys instructions found in fetched content. 4. Sanitization: No sanitization of external text is defined. This is categorized as a low-risk surface inherent to the skill's purpose.
Audit Metadata