web-fetch
Pass
Audited by Gen Agent Trust Hub on Feb 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION] (LOW): The skill documentation encourages the use of
curl,grep, andbunto fetch and process web content. These are standard tools for the skill's stated purpose of web content extraction. - [EXTERNAL_DOWNLOADS] (LOW): The skill fetches data from arbitrary user-provided URLs. While it does not download executable code from these URLs, the content is processed and displayed to the agent.
- [INDIRECT_PROMPT_INJECTION] (LOW): As a web-fetching tool, this skill has a significant attack surface for indirect prompt injection.
- Ingestion points: Web content fetched via
curland processed byhtml2markdownor the Bun script enters the agent's context. - Boundary markers: None identified in the provided documentation; content is converted to markdown and likely directly interpolated.
- Capability inventory: The skill uses
curl,grep,wc, andbun(viafetch.ts) to interact with the system and network. - Sanitization: The skill performs content extraction via CSS selectors, which acts as a basic filter, but it does not sanitize the resulting markdown text for malicious instructions.
Audit Metadata