lark-skill-maker
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides a workflow for creating new agent skills by researching API definitions from external sources (e.g., through the
lark-openapi-explorer). This creates a potential surface for indirect prompt injection where malicious content in external documentation could influence the generated skill's behavior. - Ingestion points: The workflow encourages the agent to use
lark-openapi-explorerto fetch data from the Feishu official documentation library. - Boundary markers: The provided
SKILL.mdtemplate does not include explicit boundary markers or instructions for the agent to ignore potentially malicious embedded commands in the source data. - Capability inventory: The generated skills are intended to use
lark-cli apiwhich has the capability to perform network requests (GET,POST) and modify resources via the Lark API. - Sanitization: There are no specific instructions in the
Skill Makerguidelines for sanitizing or validating the inputs derived from external documentation.
Audit Metadata