fetch-wechat-article
Pass
Audited by Gen Agent Trust Hub on Feb 20, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- PROMPT_INJECTION (LOW): Indirect Prompt Injection risk. The skill fetches untrusted article content from the web and converts it to Markdown. This content could contain hidden instructions designed to manipulate the agent's behavior.
- Ingestion points: WeChat article URLs (mp.weixin.qq.com) fetched via Playwright.
- Boundary markers: Absent. The content is saved as raw Markdown without delimiters or warnings for downstream agents.
- Capability inventory: File system writing (saveAsMarkdown), image downloading, and local script execution (node scripts/fetch.js).
- Sanitization: Performs HTML stripping, but lacks sanitization for adversarial text patterns that might influence the LLM.
- EXTERNAL_DOWNLOADS (SAFE): External dependency management. The skill requires Playwright browser binaries.
- Evidence: Documentation mentions running
npx playwright install chromiumto fetch browser dependencies. - Trust Status: Playwright is a trusted tool maintained by Microsoft; the download is downgraded to SAFE per [TRUST-SCOPE-RULE].
- COMMAND_EXECUTION (SAFE): Local command execution for scraping operations.
- Evidence: Execution of
node scripts/fetch.jsandnpxfor browser automation. - Context: These operations are legitimate and necessary for the skill's primary purpose as a scraping tool.
Audit Metadata