lark-workflow-wiki-auditor
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data retrieved from external wiki structures (such as document titles and owner metadata), which could potentially contain malicious instructions intended to influence the agent's behavior.
- Ingestion points: The skill ingests data through
lark-cli wiki spaces nodes listandlark-cli drive metas batch_queryas described in Step 2 and Step 3 ofSKILL.md. - Boundary markers: No explicit delimiters or instructions are provided to the agent to treat the retrieved wiki metadata as non-executable data or to ignore embedded instructions within titles.
- Capability inventory: The skill includes write-access capabilities, such as
lark-cli wiki +node-create(Step 6) andlark-cli wiki +move(Step 7), which could be misused if an indirect injection is successful. - Sanitization: There is no evidence of sanitization, filtering, or escaping logic applied to the data retrieved from the wiki before it is processed by the AI for health analysis or report generation.
Audit Metadata