Skillv0.1.0

ClawScan security

Weflow Group Summarizer · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignMar 11, 2026, 1:05 PM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: The skill's files, scripts, and runtime instructions are consistent with a WeFlow/WeChat group summarizer: it queries a local/ proxied WeFlow HTTP API, downloads media, saves per-group config, and prints summaries for an agent to produce; nothing in the bundle requests unrelated credentials or attempts obvious exfiltration.
Guidance: This package appears coherent for its purpose, but review these points before installing/running: 1) The proxy intentionally listens on 0.0.0.0:5032 and will expose the WeFlow API to the LAN when run — only run it in a trusted network and avoid exposing it to untrusted networks. 2) You (or the agent) will be asked to read exported member Excel files and local images; those files may contain sensitive personal data — confirm you are comfortable the agent can access and (if applicable) upload or analyze those images. 3) The Windows proxy is built locally with your Go toolchain; do not run untrusted pre-built binaries copied from unknown systems. 4) The skill does not request credentials, but it does read/write the weflow-groups.yaml you choose — keep backups if you care about that file. If you want more assurance, inspect and run the Python and Go code locally in an isolated environment before giving the agent access.

Review Dimensions

Purpose & Capability: okThe name/description (WeFlow group summarizer) aligns with the included scripts: check_weflow.py, fetch_groups.py, fetch_messages.py, convert_members.py, add_group.py and an optional weflow-proxy for remote access. Required binaries/env/config are none, matching a self-contained script-based tool. There are no unexpected credentials or unrelated dependencies.
Instruction Scope: noteSKILL.md instructs the agent/operator to run the provided Python scripts, optionally compile/run the Go reverse proxy on a Windows machine, and to read downloaded images for description. Reading saved images and the exported members .xlsx is consistent with summarization, but the instructions do require the agent/operator to access local files (images, members JSON, config YAML). The doc also tells the user to run a proxy that will expose the WeFlow API on the LAN—this is expected for remote access but raises an operational/privacy consideration (see user guidance).
Install Mechanism: okNo install spec; it's instruction-only with code files. The only build-step is a local Go cross-compile script (build-windows.sh) which copies a bundled go.mod.txt then runs 'go build'. No external download URLs or archived payloads are used. Python dependencies are minimal and listed in requirements.txt (pyyaml, openpyxl).
Credentials: okThe skill requests no environment variables, no secrets, and no config paths beyond the user-chosen weflow-groups.yaml and exported member files. All required inputs are proportional to the stated purpose.
Persistence & Privilege: okThe skill is not marked always:true and doesn't attempt to modify other skills or system-wide agent configs. It updates only the user-provided weflow-groups.yaml (expected behavior) and writes downloaded media to a user-specified images directory.