blog-factory
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external, untrusted sources such as Google Trends, Hacker News, Reddit, and Toolify to discover trending topics.
- Ingestion points: Data enters the context through discovery scripts (e.g.,
step1_discover.py) and Google Search Console API interactions. - Boundary markers: The provided instructions lack specific delimiters or instructions for the agent to ignore potentially malicious content embedded in the fetched external data.
- Capability inventory: The skill facilitates local script execution (Python and Node.js), filesystem writes to the project directory, and image generation via template scripts or external tools like Figma.
- Sanitization: No explicit sanitization or filtering steps are documented for external content before interpolation into prompts or blog files.
- [COMMAND_EXECUTION]: The skill workflow relies on executing local utility scripts and package manager commands.
- Evidence: The agent is instructed to run
gsc-review.mjs,step1_discover.py,step2_score.py,step3_produce.py, andgenerate_template_cover.py. - Context: The skill also mentions using
pnpm buildto verify the generated site andsharpfor image encoding, which are typical development tasks but involve subprocess spawning.
Audit Metadata