The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests and processes untrusted data from external sources such as Lark documents, IM messages, and events. This constitutes an indirect prompt injection surface. However, the skill mitigates this by instructing the agent to use explicit boundary markers (e.g., <fragment> and <excerpt> tags) and following strict structural parsing rules described in the lark-doc-fetch.md and lark-doc-xml.md files.
[COMMAND_EXECUTION]: The skill includes a local Python utility (lark-doc/scripts/doc_word_stat.py) used for document word counting. Static analysis of this script confirms it is benign, relying exclusively on standard Python libraries for text processing and XML parsing without any network or sensitive system access.
[EXTERNAL_DOWNLOADS]: The skill references assets and scripts from official Feishu/Lark CDN domains (e.g., feishucdn.com, feishu.cn). These are well-known, trusted services associated with the skill's primary vendor and do not pose a security risk.
[CREDENTIALS_UNSAFE]: The skill provides clear instructions for handling sensitive data like API keys and Webhook tokens. Documentation in lark-apps-openapi-key.md and lark-apps-automation.md explicitly states that these credentials are never stored by the CLI and are only displayed once to the user, promoting secure secret management practices.

lark