The Agent Skills Directory

[DATA_EXPOSURE]: The skill accesses sensitive .env files located in parent directories to retrieve API credentials.
Evidence: Scripts including lib/CalendarManager.js, check.js, and create.js use dotenv to load configurations from ../../.env or ../../../.env which contain FEISHU_APP_ID and FEISHU_APP_SECRET.
[PROMPT_INJECTION]: Indirect prompt injection surface exists through the processing of calendar event metadata.
Ingestion points: sync.js and lib/CalendarManager.js fetch event summaries and descriptions from the Feishu API.
Boundary markers: Absent; no delimiters or warnings are used when writing these strings to HEARTBEAT.md or calendar_events.json.
Capability inventory: The skill executes shell commands via Node.js scripts and has file write permissions.
Sanitization: Absent; external data is interpolated into project files used by the agent for context without escaping or validation.
[COMMAND_EXECUTION]: The skill relies on the execution of shell commands with user-influenced arguments.
Evidence: SKILL.md defines protocols for the agent to run node create.js and other scripts using parameters extracted directly from user requests.

feishu-calendar