wenqu-library
Pass
Audited by Gen Agent Trust Hub on Aug 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill facilitates the installation of external CLI tools,
open-websearchvia NPM andcrawl4aivia UV or Pip, from official package registries. These are handled as optional enhancements that require explicit user consent before installation.\n- [COMMAND_EXECUTION]: The skill's operational flow relies on shell command execution for material collection, tool setup, and health verification (e.g.,crawl4ai-doctor). It also uses browser-side JavaScript execution via Crawl4AI site recipes for handling complex pages like Sogou WeChat.\n- [PROMPT_INJECTION]: Processing and summarizing external web content introduces an attack surface for indirect prompt injection. Malicious instructions on a crawled page could attempt to influence the agent's summarization or indexing process.\n - Ingestion points: External URLs and search result metadata processed by
crwland native search tools as described inreferences/collection-playbook.md.\n - Boundary markers: No specific boundary markers or 'ignore' instructions are defined for the summarization of crawled content.\n
- Capability inventory: File system access for a local material library (
$HOME/.gogoingai/wenqu-skills/library/) and shell command execution.\n - Sanitization: The skill relies on the agent's summarization capabilities but does not specify explicit pre-processing or sanitization for raw external data inputs.
Audit Metadata