doc-fetcher

Pass

Audited by Gen Agent Trust Hub on Jul 9, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill fetches content from external documentation sites which may contain malicious instructions designed to manipulate the agent's behavior.
  • Ingestion points: Data is ingested from user-provided URLs using browser tools described in SKILL.md.
  • Boundary markers: There are no explicit markers or instructions to delimit untrusted data or ignore embedded instructions within the fetched content.
  • Capability inventory: The skill utilizes browser tools for data collection and performs local file system writes to store the retrieved markdown files.
  • Sanitization: The skill performs content cleaning to remove boilerplate (navigation, footers, ads), but does not implement security sanitization specifically against prompt injection techniques.
  • [EXTERNAL_DOWNLOADS]: Fetches content from external web sources. It references the Cloudflare Browser MCP server as a prerequisite for its functionality, which is a well-known service for browser-based automation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 9, 2026, 09:52 PM
Security Audit — agent-trust-hub — doc-fetcher