web-fetch-to-markdown
Pass
Audited by Gen Agent Trust Hub on May 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a bash wrapper (
scripts/fetchmd) to execute a local Node.js script (fetchmd.js) for content processing and conversion. - [DATA_EXFILTRATION]: The skill performs network operations to fetch content from external URLs.
- Ingestion points: External content is fetched from user-provided URLs via the
fetchmdscript. - Boundary markers: The instructions in
SKILL.mddefine protocol boundaries (onlyhttp://orhttps://), though technical delimiters for the fetched content are not specified in the provided files. - Capability inventory: The script uses
nodeto process external data. - Sanitization: The tool converts HTML to Markdown, providing structural normalization of untrusted web content before it is presented to the agent.
- [PROMPT_INJECTION]: Because the skill fetches and processes data from arbitrary websites, it is subject to indirect prompt injection. Maliciously crafted content on a fetched page could include hidden instructions intended to override the agent's behavior or exfiltrate data from the current session.
Audit Metadata