The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the lark-channel-bridge package from a public registry (npm or pnpm). This package is not associated with a recognized trusted organization, representing the introduction of unverified external code.
[COMMAND_EXECUTION]: The bridge's primary function is to receive messages from the Feishu/Lark messenger and forward them as executable tasks to local coding agents (claude or codex). This creates a remote command execution interface accessible from a public chat platform.
[PERSISTENCE]: The bridge provides commands (lark-channel-bridge start) to automatically register itself as a background service using system-level tools such as systemd on Linux, launchd on macOS, and Task Scheduler on Windows, ensuring the bridge remains active across reboots.
[PRIVILEGE_ESCALATION]: The skill documents a full access mode that explicitly configures the underlying agents to use high-privilege flags (bypassPermissions for Claude and danger-full-access for Codex). This grants the remote messenger interface the ability to use all tools, perform authentication flows, and write to any file system location accessible to the agent.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted remote data, making it highly susceptible to indirect prompt injection attacks.
Ingestion points: Direct messages, group mentions, and file attachments received from the Feishu/Lark messenger platform.
Boundary markers: The instructions do not define delimiters or warnings to prevent the agent from obeying instructions embedded within the remote messages.
Capability inventory: The bridge provides access to all capabilities of the local coding agents, including shell command execution and file system modifications.
Sanitization: There is no evidence of input sanitization or validation of the remote content before it is processed by the agent.
[CREDENTIALS_UNSAFE]: The tool includes a command (profile export --include-secrets) that allows for the extraction of sensitive configuration data, including encrypted app secrets, which could be exploited if an attacker gains control of the session.

lark-coding-agent-bridge