docs-workflows
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted metadata from external sources (Bilibili and Juejin) to automate the addition of tutorials. While relevance filtering and manual review steps are defined, the agent could be exposed to malicious instructions embedded in external titles or descriptions. * Ingestion points: Bilibili API fields (data.title, data.pic) and Juejin scraping. * Boundary markers: Explicit relevance filters with inclusion/exclusion keywords are provided. * Capability inventory: File writes to TutorialsGrid.tsx, manageStorage uploads, and downloadRemoteFile. * Sanitization: Relies on agent filtering and user picking.
- [COMMAND_EXECUTION]: Instructions guide the agent to run local maintenance scripts using Node.js, such as scripts/build-compat-config.mjs and scripts/sync-config.mjs. These scripts are internal to the repository and are used for build and verification tasks.
- [EXTERNAL_DOWNLOADS]: The skill automates downloading external assets, including Bilibili video thumbnails and IDE icons from official websites or the lobe-icons GitHub repository. These assets are processed and uploaded to the project's cloud storage environment.
Audit Metadata