feishu-doc-workflow
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill fetches content from external Feishu/Lark URLs, allowing untrusted data to enter the agent context.\n
- Ingestion points: External document content is retrieved via
lark-cli docs +fetch(SKILL.md).\n - Boundary markers: There are no explicit instructions to use delimiters or warnings to ignore instructions embedded in the retrieved documents.\n
- Capability inventory: The skill possesses writing capabilities to Feishu/Lark docs, API access, and the ability to execute local conversion scripts (
mineru-md.sh).\n - Sanitization: There is no evidence of sanitization or validation of the input data before it is processed or written back.\n- [COMMAND_EXECUTION]: The skill instructs the agent to execute specific local binaries and shell scripts.\n
- Evidence: Commands involving
~/.local/bin/lark-cliand a local PDF processing script at/Users/wangpu/Library/Mobile Documents/iCloud~md~obsidian/Documents/WorldModelVault/.tools/mineru-md.share documented in SKILL.md.\n- [CREDENTIALS_UNSAFE]: The skill references a sensitive configuration file path that contains authentication secrets.\n - Evidence: SKILL.md provides instructions for inspecting
~/.lark-cli/config.json, although it includes a safety directive to mask secrets during inspection.
Audit Metadata