content-brief-factory

Pass

Audited by Gen Agent Trust Hub on Apr 12, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes local Python scripts (e.g., site-content-catalog/scripts/catalog_site.py, review-scraper/scripts/scrape_reviews.py, reddit-scraper/scripts/scrape_reddit.py) to perform site auditing and data scraping tasks.
  • [EXTERNAL_DOWNLOADS]: The skill connects to multiple external API providers (SerpAPI, Serper.dev, DataForSEO, ValueSERP) and fetches content from public websites and social platforms like Reddit to inform the content briefs.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core functionality of processing external data.
  • Ingestion points: Untrusted data is ingested from external webpages, Reddit posts, and review platforms (G2, Capterra) through scraping tools.
  • Boundary markers: The instructions lack explicit boundary markers or system instructions to prevent the agent from following instructions potentially embedded in the scraped content.
  • Capability inventory: The agent has the capability to execute shell commands (python3) and perform file system writes (clients/.../content/briefs/).
  • Sanitization: No validation or sanitization of the scraped external content is described before it is used to generate the final content brief output.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 12, 2026, 02:42 PM