wechat-channels-crawler
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPERSISTENCE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes content (titles, author names) from an external API (redfox.hk) and displays it to the user. \n
- Ingestion points: The
searchWorkendpoint (viascripts/search_wechat_channels.py) returns user-generated content from WeChat Channels. \n - Boundary markers: While the data is formatted into a Markdown table, there are no explicit instructions for the agent to ignore or sanitize embedded instructions within the work titles or descriptions. \n
- Capability inventory: The agent can write to the UI, execute shell commands for payment handling, and create scheduled tasks via
qoder_cron. \n - Sanitization: No sanitization of the external text content is performed beyond character length truncation for display formatting. \n- [EXTERNAL_DOWNLOADS]: The skill uses
npxto download and install the@alipay/agent-paymentpackage from a well-known service (Alipay) to handle credit recharges. \n- [COMMAND_EXECUTION]: The skill executes local Python scripts (search_wechat_channels.py) and specialized CLI tools (alipay-bot) for search and payment orchestration. \n- [PERSISTENCE]: The skill utilizes theqoder_crontool to create scheduled tasks for daily content pushes. This is a documented feature triggered only upon user confirmation via an interactive prompt.
Audit Metadata