ai-marketing-skills-tom-babb
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The 'Knowledge Scrape to Content' workflow describes a process for gathering external data from public forums like Reddit and feeding it into an LLM to generate content briefs. This pattern introduces a vulnerability surface where malicious instructions embedded in scraped comments could attempt to influence the LLM's output.
- Ingestion points: The
scrape_pain_pointsfunction (via Reddit API) and the synthesis prompt template inSKILL.mdingestion points for untrusted data. - Boundary markers: The synthesis prompt uses a simple
[paste scraped pain points]placeholder without robust delimiters or specific instructions for the agent to ignore embedded commands. - Capability inventory: The workflow targets content generation (outlines, blog posts) rather than sensitive tool execution or file system access.
- Sanitization: No input sanitization or filtering is implemented in the provided Python examples or prompt templates to strip potential prompt injection payloads from the scraped text.
Audit Metadata