The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the synchronization logic in sync.js. External data fetched from the Feishu (Lark) API is written into the HEARTBEAT.md file, which may influence the agent's future behavior if it contains malicious instructions.\n
Ingestion points: sync.js retrieves event summaries and descriptions from the Feishu API via the CalendarManager class.\n
Boundary markers: No delimiters or protective instructions are used when interpolating external event data into HEARTBEAT.md.\n
Capability inventory: The skill can write to the local file system and manage calendar events (list, create, and delete) using the provided Feishu credentials.\n
Sanitization: The skill does not sanitize or escape the event summaries before writing them to the project's heartbeat file.\n- [DATA_EXFILTRATION]: The skill performs file system operations on sensitive files and directories located outside of the skill's own folder. lib/CalendarManager.js attempts to load environment variables from ../../../.env, while sync.js modifies ../../HEARTBEAT.md and ../../memory/calendar_events.json.\n- [DATA_EXFILTRATION]: A discrepancy was detected between the security guardrails described in SKILL.md and the actual capabilities of the included scripts. While the documentation states that the skill only supports listing, searching, checking, syncing, and creating events, the scripts cleanup.js and setup_routine.js contain functional code to delete events from the Feishu calendar.

feishu-calendar-hardened