geo-competitor-scanner

Pass

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill retrieves public metadata and technical files (e.g., /llms.txt, /robots.txt) from domains provided by the user. These requests use the standard requests library with a 10-second timeout and are essential for the skill's analytical function.
  • [COMMAND_EXECUTION]: The skill uses local Python scripts to perform data processing. These scripts are implemented using standard libraries and do not execute external commands or arbitrary user strings.
  • [REMOTE_CODE_EXECUTION]: No remote code execution vectors were identified. The skill parses HTML and JSON content using robust libraries (BeautifulSoup and the native json module) that do not evaluate the content as code.
  • [DATA_EXPOSURE]: There is no access to sensitive system files or credentials. The skill only processes publicly accessible information found on the targeted websites.
  • [PROMPT_INJECTION]: The risk of indirect prompt injection is mitigated by the skill's design.
  • Ingestion points: scripts/scan_competitors.py fetches external website content via requests.get().
  • Boundary markers: Scan results are structured into a markdown comparison report.
  • Capability inventory: Network access (requests), HTML parsing (BeautifulSoup), and JSON parsing.
  • Sanitization: The skill extracts specific technical metrics (counts of headers, FAQ mentions, schema types) rather than raw text ingestion, preventing accidental interpretation of external content as instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 5, 2026, 02:33 AM