feishu-smart-doc-writer

Warn

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DATA_EXFILTRATION]: The skill performs sensitive file access to retrieve credentials independently of the platform's provided context.
  • The file feishu_smart_doc_writer.py contains logic in the _get_tenant_access_token method that attempts to read from ~/.openclaw/openclaw.json and ~/.openclaw/config.json.
  • It specifically extracts appId and appSecret from these files to facilitate authentication.
  • While the credentials are used for legitimate communication with the official Feishu API (open.feishu.cn), accessing global platform configuration files is a high-risk behavior as these files typically contain secrets for multiple services and channels beyond the scope of this specific skill.
  • [CREDENTIALS_UNSAFE]: Hardcoded identifiers are present in the skill's default configuration files.
  • The files config.json and user_config.json both contain a hardcoded Feishu OpenID (ou_5b921cba0fd6e7c885276a02d730ec19).
  • Although this is a user identifier rather than a secret key, including specific test identifiers in distributed configuration files is a poor security practice.
  • [INDIRECT_PROMPT_INJECTION]: The skill implements a local document indexing system that creates a surface for indirect instruction injection.
  • Ingestion points: The search_docs and list_docs tools in index_manager.py read metadata from a local markdown file ~/.openclaw/workspace/memory/feishu-docs-index.md.
  • Boundary markers: None identified; the markdown table rows are parsed into structured data and returned directly to the agent's context.
  • Capability inventory: The skill can write to the local file system, perform network operations via aiohttp, and manage document permissions through the Feishu API.
  • Sanitization: The indexing logic extracts the first 100 characters of written content to create document summaries without sanitization. If a document's title or content (which may be sourced from untrusted third parties) contains malicious instructions, they could be ingested into the agent's prompt during search or list operations.
  • [EXTERNAL_DOWNLOADS]: The skill communicates with the official Feishu/Lark platform.
  • It makes network requests to https://open.feishu.cn for authentication token retrieval and document ownership transfer.
  • These operations target a well-known service and are consistent with the skill's stated purpose of document management.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 19, 2026, 11:32 PM
Security Audit — agent-trust-hub — feishu-smart-doc-writer