The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill setup involves installing the @larksuite/cli package from the official npm registry. This is the official developer tool provided by the platform vendor (Lark/Feishu).
[COMMAND_EXECUTION]: The scripts/lark_setup.py script uses the Python subprocess module to call the lark-cli tool for configuration. The script correctly passes arguments as a list rather than using a shell string, preventing command injection. Additionally, the skill guides the agent to use lark-cli via the Bash tool for its primary operations, which is intended behavior.
[PROMPT_INJECTION]: The skill processes external data (Lark messages, emails, and cloud documents) which creates an attack surface for indirect prompt injection. Malicious instructions embedded in these external sources could theoretically attempt to influence the agent's logic.
Ingestion points: Documented commands such as +chat-messages-list, +messages-search, and +docs-get in SKILL.md and reference files.
Boundary markers: No specific delimiters or boundary markers are instructed for the agent when interpolating this external content into prompts.
Capability inventory: The skill possesses significant capabilities via the Bash tool, including the ability to send messages, delete files, and update database records across the Lark platform.
Sanitization: The documentation in references/lark-mail.md explicitly warns that email content should be treated as untrusted input and recommends sanitization, though no automated sanitization script is provided within the skill itself.

lark-unified