web-crawl
Pass
Audited by Gen Agent Trust Hub on Jul 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes a local Python script
harness/scripts/web-crawl.pyto handle web fetching, markdown conversion, and self-testing. This is a standard execution pattern for skill-specific logic. - [EXTERNAL_DOWNLOADS]: The skill is designed to fetch content from external URLs using the
urlliblibrary in its builtin backend. It also references well-known external services like Firecrawl and Crawl4AI for advanced functionality. - [PROMPT_INJECTION]: The skill processes untrusted data from the web, which introduces an indirect prompt injection surface.
- Ingestion points: The
fetchcommand ingests content from arbitrary external URLs, and themdcommand processes local HTML files. - Boundary markers: The skill encourages distilling content into markdown to reduce token usage and noise, which acts as a structural boundary, though explicit instructions to ignore embedded commands are not mentioned.
- Capability inventory: The skill executes a local Python script and has the ability to write distilled content to the local file system (e.g., the
raw/directory). - Sanitization: The builtin backend uses regex to convert HTML to markdown, which naturally strips out executable tags like
<script>and<object>, providing a layer of sanitization for the ingested content.
Audit Metadata