juejin-smart-publish
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECREDENTIALS_UNSAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The script
scripts/publish.pystores sensitive Juejin session cookies in a local file at~/.juejin_cookies/cookies.jsonto maintain authentication state. - [COMMAND_EXECUTION]: The skill executes a Python script
scripts/publish.pythat leverages the Playwright library to automate browser interactions, including navigating to external URLs and filling forms. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted Markdown data and command-line arguments to populate article fields and API payloads.
- Ingestion points:
scripts/publish.pyreads user content from the--content-file,--title, and--abstractarguments. - Boundary markers: Absent; the skill does not use delimiters or safety instructions to isolate user content from the automation logic.
- Capability inventory:
scripts/publish.pypossesses file system access (writing cookies and screenshots), network access to Juejin domains, and browser automation capabilities. - Sanitization: Absent; input data is directly passed to browser fields and API endpoints without validation.
Audit Metadata