Skill Taxonomy Router Pro

Route user tasks to the most relevant skills using a layered taxonomy, risk model, and minimum-necessary-loading strategy. Use when deciding which skill to l...

This skill is largely coherent with its stated router/classifier purpose, but it bundles many executable scripts that can download other skills, alter local indexes, and delete directories. Before installing or executing anything from this package: 1) Inspect the scripts (intake_skill.py, cleanup_download_inbox.py, detect_skill_overlap.py, etc.) and search for absolute paths and subprocess invocations. 2) Do not run these scripts as-is on a production machine — they assume developer paths (/home/parallels/...) and may perform destructive actions. 3) If you need the automation, adapt paths and add safeguards (dry-run flags, explicit confirmation prompts, sandboxed workdir) and run inside an isolated environment or VM. 4) Verify what 'clawhub install' will fetch (source/trust) before using intake_skill.py. 5) Prefer using the SKILL.md policy and manual classification steps for initial use; only run automated scripts after you understand and harden them. If you want, I can list the exact lines that invoke downloads, deletions, and absolute paths so you can review them quickly.

Type: OpenClaw Skill Name: skill-taxonomy-router-pro Version: 1.1.0 The skill bundle is a governance and organization tool designed to classify, route, and assess the risk of other skills within the OpenClaw ecosystem. It implements a structured taxonomy (A–Z domains) and a risk model (R0–R4) to ensure the agent uses the 'minimum necessary' capabilities for a task. While the bundle includes scripts that perform file system operations (e.g., `shutil.rmtree` in `cleanup_download_inbox.py` and `subprocess.run` for `clawhub install` in `intake_skill.py`), these are strictly scoped to a designated 'inbox' directory and are transparently documented as part of the maintenance workflow. The instructions in `SKILL.md` explicitly mandate user approval for risky operations and prioritize low-risk paths, demonstrating a security-conscious design without evidence of malicious intent or data exfiltration.