feishu-claude-code-bridge
Pass
Audited by Gen Agent Trust Hub on Jun 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The bridge executes the
claudeCLI locally to process messages received from Feishu. It also manages background processes using platform-specific service managers like systemd, launchd, and Windows Task Scheduler.\n- [EXTERNAL_DOWNLOADS]: The skill requires the installation of thelark-channel-bridgepackage from the NPM registry. This is a vendor-provided tool necessary for the bridge's operation.\n- [DATA_EXFILTRATION]: By design, the skill forwards local chat session data and downloaded media files to Feishu/Lark servers to facilitate the remote chat interface.\n- [PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by ingesting untrusted messages from Feishu and passing them to the local Claude agent.\n - Ingestion points: Feishu/Lark messenger chat messages forwarded to the bridge.\n
- Boundary markers: None explicitly implemented at the bridge level; messages are forwarded verbatim.\n
- Capability inventory: Access to the local
claudeCLI, which can read/write files and execute commands within workspaces.\n - Sanitization: Includes configurable access control lists (ACLs) to restrict the bot to specific users or groups and protects administrative commands.
Audit Metadata