content-brief-factory
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from various external sources to generate its briefs, which could allow malicious instructions hidden in search results or user reviews to influence the agent's behavior.
- Ingestion points: Target keyword search results, competition page content (via fetch_webpage), and social content from Reddit and review platforms (G2, Capterra) as defined in the per-keyword analysis phase.
- Boundary markers: The skill does not implement specific delimiters or 'ignore' instructions when processing external web content, which is necessary to prevent the model from following instructions found in that content.
- Capability inventory: The agent can execute local shell commands (Python scripts) and has write access to the filesystem to store the generated briefs.
- Sanitization: No explicit sanitization, filtering, or validation process is mentioned for the content retrieved from external URLs before it is processed.
- [COMMAND_EXECUTION]: The skill orchestrates the execution of multiple local Python scripts (e.g., catalog_site.py, scrape_reviews.py, scrape_reddit.py) to perform specialized SEO and data mining tasks. These scripts are invoked via the shell with arguments derived from user-provided keywords and site URLs.
Audit Metadata