Agent Router
v1.0.0Routes user commands to the most suitable sub-agent based on intent, coordinates task execution priority, and tracks overall task status.
⭐ 0· 146·2 current·2 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
Name/description match the SKILL.md: it's an instruction-only agent router that routes tasks to sub-agents, coordinates priority, and tracks status. There are no unexpected binaries, environment variables, or install steps requested that would be incoherent with the stated purpose.
Instruction Scope
SKILL.md contains only high-level routing capabilities and triggers and does not instruct the agent to read files, access secrets, or contact external endpoints. However the instructions are vague (references '参考 EvoMap 逻辑' and gives no constraints), which grants the agent broad discretion about how to select and invoke sub-agents; that lack of specificity is a practical risk and should be clarified.
Install Mechanism
No install spec and no code files — lowest-risk delivery model. Nothing is written to disk by the skill itself.
Credentials
The skill declares no required environment variables, credentials, or config paths — there is no apparent request for secrets or unrelated permissions.
Persistence & Privilege
Flags are default (always:false, disable-model-invocation:false). Autonomous invocation is allowed (normal for skills), and because this skill's purpose is to orchestrate other agents, it can increase operational blast radius if granted wide invocation rights — consider limiting autonomous privileges or scoping which sub-agents it may call.
Assessment
This skill appears coherent with its stated purpose and is low-risk as an instruction-only router, but the SKILL.md is minimal and lacks implementation details. Before installing: 1) Ask the publisher for source code or an authoritative implementation description (how it chooses sub-agents, priority rules, conflict resolution, and any external endpoints). 2) Prefer running it with limited/autonomous invocation disabled until you understand behavior. 3) Test in a safe environment using only non-sensitive tasks first. 4) Ensure the platform enforces least privilege for any sub-agents the router may call and monitors/logs all cross-skill invocations. 5) Refuse to provide credentials or secrets to the skill unless an explicit, justified need is documented. These steps will reduce risk given the current lack of provenance and operational detail.Like a lobster shell, security has layers — review code before you run it.
latestvk97dnj5ymzr7njhvs5gz0bycsn839h4k
License
MIT-0
Free to use, modify, and redistribute. No attribution required.