The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources which could potentially contain malicious instructions.
Ingestion points: Fetches content from cloud documents (lark-doc-fetch.md), retrieves chat messages (LarkMessageTrigger in lark-base-workflow-schema.md), and reads database records (lark-base-record-query-and-analysis-sop.md).
Boundary markers: The instructions lack explicit boundary markers or warnings to disregard instructions embedded within the ingested data.
Capability inventory: The agent can perform significant actions based on processed data, including executing HTTP requests (HTTPClientAction), modifying databases (SetRecordAction), and approving workflows (lark-approval-tasks-approve.md).
Sanitization: There is no evidence of filtering or sanitization of external content before it is used in decision-making or command construction.
[DYNAMIC_EXECUTION]: The skill facilitates the generation and execution of code at runtime to support its design and development features.
Javascript Evaluation: The fork-verifier-agent.md sub-agent is instructed to use eval_js to inspect and verify UI layouts in a served environment.
SQL Execution: The +db-execute command in lark-apps-db-execute.md allows for the execution of arbitrary SQL strings against application databases.
Runtime Compilation: The creative-design.md module generates React and Babel code that is executed within a browser-based preview environment.
[EXTERNAL_DOWNLOADS]: The skill references and installs software from external repositories.
Fetches the @larksuite/cli package from the official NPM registry for CLI updates.
Loads several Javascript libraries (React, ReactDOM, Babel, ECharts) from ByteDance's official CDN (sf3-scmcdn-cn.feishucdn.com).
Loads fonts from a self-hosted mirror on miaoda.feishu.cn.

lark