The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a local binary lark-cli to perform task management operations. All write-intensive commands across the reference files (e.g., lark-task-create.md, lark-task-update.md) include explicit 'CAUTION' blocks requiring the agent to obtain user confirmation before proceeding.
[SAFE]: No malicious patterns such as credential theft, hardcoded secrets, or multi-layer obfuscation were detected. The skill correctly identifies and handles sensitive user identity parameters, specifically instructing the agent to resolve open_id for 'me' requests using official methods.
[SAFE]: Analysis of indirect prompt injection surfaces:
Ingestion points: Data is ingested via lark-cli task list, get, and +get-my-tasks commands which fetch task summaries and descriptions (documented in SKILL.md and lark-task-get-my-tasks.md).
Boundary markers: Absent. The instructions do not specify the use of delimiters when presenting or processing retrieved task content.
Capability inventory: The skill possesses extensive write capabilities, including task creation, updating, and member assignment (detailed in lark-task-create.md, lark-task-update.md, and lark-task-assign.md).
Sanitization: Absent. There is no specific instruction to sanitize or escape task content before it is interpolated into prompts.
Evaluation: While an ingestion surface exists for indirect prompt injection via task content, the potential impact is mitigated by the mandatory requirement for user confirmation on all state-changing commands.

lark-task