The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/recall.py uses os.execv to execute the resource-memory tool via the uv runner. This is used to wrap the tool's CLI and allows the agent to perform local database operations.
[COMMAND_EXECUTION]: The skill includes management scripts like scripts/wechat_data/doctor.py and scripts/wechat_data/start_metrics_service.py that execute shell commands (e.g., networksetup) and launch local services via subprocess calls.
[EXTERNAL_DOWNLOADS]: The article acquisition logic in scripts/wechat_data/download_urls.py sends user-provided WeChat URLs to an external API (https://down.mptext.top) to retrieve content. This is a documented external dependency used for scraping public articles.
[PROMPT_INJECTION]: The skill ingests external content from the web and WeChat articles into a local knowledge base. This creates an indirect prompt injection attack surface where malicious text within a saved article could influence the agent's future responses or actions when the resource is recalled.

recall-resources