Back to skill
Skillv1.0.0

ClawScan security

agent-creator-feishu · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignApr 29, 2026, 2:27 AM
Verdict
benign
Confidence
high
Model
gpt-5-mini
Summary
The skill's requests and instructions are coherent with its stated purpose: it guides creating an OpenClaw agent and binding it to a Feishu group by running local openclaw CLI commands and editing local OpenClaw configuration files.
Guidance
This skill is internally consistent and simply guides local OpenClaw operations (creating workspaces, writing agent files, updating ~/.openclaw/openclaw.json, and binding to Feishu groups). Before using it: back up your ~/.openclaw/openclaw.json and any important workspace files; confirm your Feishu gateway/credentials are already configured (the skill does not request Feishu tokens); verify the administrator ID it will record is correct (it takes sender_id from the current message); and only run the suggested shell commands if you trust the environment where they will modify files. If you want extra safety, inspect the generated IDENTITY.md / SOUL.md / AGENTS.md / SKILL.md before applying them and test bindings in a non-production group first.

Review Dimensions

Purpose & Capability
okName/description (create an agent and bind to Feishu) match the actual instructions: use the openclaw CLI, write agent files into ~/.openclaw workspaces, and update bindings/whitelists. No unrelated credentials, binaries, or external services are requested.
Instruction Scope
okSKILL.md stays on‑topic: it collects user requirements, generates agent config files, runs openclaw agents add/bind, and instructs editing ~/.openclaw/openclaw.json and workspace files. It references reading sender_id from the current message and checking logs in /tmp — these are relevant to provisioning and verification. The instructions do not ask to exfiltrate data or call unknown external endpoints.
Install Mechanism
okNo install spec and no code files — instruction-only skill. No downloads or archive extraction are performed or recommended, minimizing install-time risk.
Credentials
okThe skill declares no required environment variables or credentials. It operates on local OpenClaw files and uses the current message context (sender_id) which is appropriate for assigning admin privileges.
Persistence & Privilege
noteThe guide instructs modifying the global OpenClaw config (~/.openclaw/openclaw.json) and adding bindings, which is expected for binding agents but does require write access to user configuration. This is a legitimate need for the feature but is a privileged action — users should be aware it changes system config files.