← Back to Skills Marketplace
Clawrank
by
JoeyCacciatore3
· GitHub ↗
· v1.0.0
· MIT-0
99
Downloads
1
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install nate-clawrank
Description
Agent performance scoring system for OpenClaw agents. 7 dimensions scored 0-10, crab-themed tiers, evidence-based, with trajectory tracking. Use at session e...
Usage Guidance
This is an instruction-only scoring rubric and on its face is benign, but there are two things to check before enabling it broadly: (1) the SKILL.md says it "Integrates with agent-sync" but provides no endpoint, API, or credential instructions — ask the author how agent-sync integration works and whether it will send conversation content anywhere; (2) the rules require a one-line evidence entry per dimension, which means the agent will use session context (chat history, user prompts) to produce those lines — confirm you’re comfortable with that data being used/stored and whether peer review will transmit it externally. If you are unsure, test the skill in a restricted/sandboxed agent (or disable autonomous invocation) and verify where any peer-review data is posted and what auth it uses.
Capability Analysis
Type: OpenClaw Skill
Name: nate-clawrank
Version: 1.0.0
The skill bundle is a purely instructional framework for an AI agent to self-evaluate its performance using a rubric called 'ClawRank'. It contains no executable code, network requests, or file system access, and the instructions in SKILL.md are limited to scoring logic and performance tiering without any malicious prompt injection or data exfiltration triggers.
Capability Assessment
Purpose & Capability
The name, description, and instructions all align: this is a performance-scoring rubric for agents. However, the SKILL.md claims it "Integrates with agent-sync for peer review" but the skill declares no required env vars, endpoints, or instructions for how to access agent-sync. That integration claim is unexplained and may be a missing or incomplete dependency.
Instruction Scope
Instructions are tightly scoped to scoring seven dimensions, formatting the report, and tracking trajectory. They require the agent to produce short evidence lines per dimension which will naturally use session context (conversation history). The spec does not explicitly instruct reading unrelated system files or secrets, but it does hint at posting/recording to agent-sync without describing what data is sent or how — this ambiguity could lead to unexpected transmission of session content.
Install Mechanism
Instruction-only skill with no install steps and no code files. That reduces surface area — nothing is downloaded or written to disk by the skill itself.
Credentials
The skill requests no environment variables or credentials, which is consistent with a local rubric. However, because it references agent-sync peer review, the lack of any declared credentials or endpoint is odd: if agent-sync requires auth, that requirement is not documented here.
Persistence & Privilege
always is false (normal) and the skill does not request persistent configuration or system-wide changes. Autonomous invocation is allowed by default — expected for skills — but consider whether you want the agent to trigger self-evaluations without explicit user consent.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install nate-clawrank - After installation, invoke the skill by name or use
/nate-clawrank - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of ClawRank: a comprehensive agent performance scoring system for OpenClaw agents.
- Introduces 7 evidence-based scoring dimensions (0–10 each), normalized to a 100-point scale.
- Crab-themed performance tiers (Frozen to King Crab) with defined ranges and meanings.
- Standardized session-ending format for score reports, including evidence for each dimension.
- Weekly trajectory tracking for progress monitoring.
- Peer review integration via agent-sync; enforces explanation on major self/peer disagreement.
- Strict rules for scoring consistency, evidence, and trend focus.
Metadata
Frequently Asked Questions
What is Clawrank?
Agent performance scoring system for OpenClaw agents. 7 dimensions scored 0-10, crab-themed tiers, evidence-based, with trajectory tracking. Use at session e... It is an AI Agent Skill for Claude Code / OpenClaw, with 99 downloads so far.
How do I install Clawrank?
Run "/install nate-clawrank" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Clawrank free?
Yes, Clawrank is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Clawrank support?
Clawrank is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Clawrank?
It is built and maintained by JoeyCacciatore3 (@joeycacciatore3); the current version is v1.0.0.
More Skills