The Agent Skills Directory

[PROMPT_INJECTION]: The skill introduces an indirect prompt injection surface as it is designed to ingest and process messages from the external Feishu platform.
Ingestion points: The handleMessage function in add/src/channels/feishu.ts receives data via the im.message.receive_v1 WebSocket event.
Boundary markers: The implementation does not include specific delimiters or instructions for the agent to ignore potentially malicious instructions embedded in user messages.
Capability inventory: The system architecture (modify/src/index.ts) allows the agent to execute code within containers and interact with multiple messaging channels.
Sanitization: While message content is parsed from JSON, the raw text content is passed into the agent's processing pipeline without specific sanitization or filtering.
[COMMAND_EXECUTION]: The documented setup process requires the execution of system commands for environment configuration and service lifecycle management.
Evidence: SKILL.md includes instructions for building the project with npm run build and restarting background services using launchctl or systemctl.
[EXTERNAL_DOWNLOADS]: The skill utilizes official external libraries and connects to verified service endpoints.
Evidence: It installs the @larksuiteoapi/node-sdk package and communicates with the official Feishu API at open.feishu.cn.
[CREDENTIALS_UNSAFE]: The skill manages sensitive application credentials required for the Feishu integration.
Evidence: The skill configuration involves setting FEISHU_APP_ID and FEISHU_APP_SECRET in the .env file, which are then accessed by the application logic.

add-feishu