lark-openapi-explorer
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill fetches documentation from external URLs (open.feishu.cn and open.larksuite.com) using WebFetch. This content is parsed to identify API endpoints, parameters, and calling conventions, which could allow malicious instructions in compromised documentation to influence agent behavior.
- Ingestion points: SKILL.md (via WebFetch calls to Feishu/Lark documentation indices and individual API pages).
- Boundary markers: No specific delimiters or "ignore instructions" markers are used when processing the fetched documentation content.
- Capability inventory: The skill executes shell commands via
lark-cli api(GET, POST, PUT, DELETE, PATCH). - Sanitization: The skill relies on manual safety rules requiring user confirmation for write/delete operations, but lacks programmatic sanitization of the fetched content.
- [COMMAND_EXECUTION]: The skill dynamically constructs and executes shell commands using the
lark-clitool. The commands are built based on specifications extracted from external web resources. - [EXTERNAL_DOWNLOADS]: The skill fetches content from
open.feishu.cnandopen.larksuite.com. These are official developer portals for the service (Feishu/Lark) that the skill is designed to manage.
Audit Metadata