audit-website-aeo

Pass

Audited by Gen Agent Trust Hub on May 19, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is subject to indirect prompt injection as it ingests untrusted data from external websites.
  • Ingestion points: External website content (titles, meta tags, and body text) is crawled by scripts/aeo-audit.mjs and stored in aeo-audit.json for agent review.
  • Boundary markers: SKILL.md lacks explicit boundary markers or instructions for the agent to ignore potential instructions embedded within the textExcerpt or metadata fields.
  • Capability inventory: The skill utilizes subprocess execution (Node.js) and file read/write operations.
  • Sanitization: The provided crawler script performs basic HTML tag stripping and entity decoding but does not filter content for adversarial instructional patterns.
  • [COMMAND_EXECUTION]: The skill executes a local Node.js script using parameters derived from user input (URL and output path).
  • Evidence: node <skill-path>/scripts/aeo-audit.mjs <url> --max-pages=10 --out=<output-dir>/aeo-audit.json in SKILL.md.
  • Risk: This creates a surface for command injection if the agent does not properly sanitize the URL or directory path before passing them to the shell. This is a common design pattern for CLI-interfacing skills.
Audit Metadata
Risk Level
SAFE
Analyzed
May 19, 2026, 12:29 PM
Security Audit — agent-trust-hub — audit-website-aeo