Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
Name/description (task routing) align with the SKILL.md routing tables and logic. It does not request unrelated binaries, env vars, or config paths.
Instruction Scope
SKILL.md limits behavior to parsing task text, extracting keywords (mentions TF‑IDF) and routing to named Agents/Skills. However it is implementation‑light (no detail how TF‑IDF is computed, thresholds, or tie‑breaking) and does not mention handling of sensitive data or asking user consent before forwarding tasks to other agents.
Install Mechanism
Instruction-only skill with no install spec and no code files — nothing is written to disk or downloaded.
Credentials
No environment variables, credentials, or config paths are requested; requested scope is minimal and proportional to the described function.
Persistence & Privilege
always:false and autonomous invocation is allowed (normal). The SKILL.md states '自动触发' — automatic routing is expected for a router, but automatic forwarding of user messages to other agents can increase blast radius and privacy risk; the skill does not document consent/visibility controls.
Assessment
This skill is coherent with its stated purpose, but before installing consider: (1) routing forwards user messages to other agents/skills — ensure you trust those targets and understand their data handling; (2) ask the author how ambiguous tasks, tie breaks, and sensitive data (PII, credentials) are handled — prefer explicit consent or allowlist/blacklist controls; (3) test in a safe environment to verify routing behavior and fallback handling. If you need stronger privacy guarantees, request the developer add explicit redaction/consent steps and deterministic routing thresholds.Like a lobster shell, security has layers — review code before you run it.
latestvk97d30kzs5rj6w163d6qp5wnk9844t56
License
MIT-0
Free to use, modify, and redistribute. No attribution required.