MMCPMarkets
AS

andrewnggirl/skilllens

Developer tools

This page was translated automatically. Technical details may differ from the original. View the original

44 stars Quality 60 Trend 16MIT
Repository Install

Open-source self-hosted web tool for evaluating Agent Skills with rubric scores, Deep Review, and improvement suggestions.

Overview

https://github.com/user-attachments/assets/8bc9bfce-bfe5-4c3a-a915-9aaf1520969e SkillLens is an open-source, self-hosted web tool for evaluation. You can upload a SKILL.md, a skill folder, or a skill compressed package. SkillLens generates a quantitative report, including scoring criteria, evidence, SkillLens Deep Review, market signals, and actionable improvement suggestions. It is aimed at developers building skills for Cursor, Claude, OpenClaw, or similar Agent ecosystems, helping you answer a core question: - (Review Perspective) - : 5 major pillars, 100-point scale; Enable differentiated sub-dimensions by skill structure (atomic / pipeline / composite) - (stackable): In addition to the general total score, additional domainExpert scores are output for financial scenarios, covering investment and financing, quantitative trading, stock trading, securities, banking, investment education, etc - (Evaluation Depth) - : Deterministic checks (pass / partial / fail / not_applicable), returns in seconds, without consuming LLM quotas - : Above the rule score, the LLM evaluates subjective dimensions; Rule items are still determined by the official scorer to prevent LLM from acting freely - (Call entry) - : For users — browser drag-and-drop upload, visual report - : For code agents (such as Cursor / WorkBuddy / Hermes) ——-- agent-wizard to guide selection, and deepReviewCertificate verification, prevents agents from fabricating scores The three dimensions can be freely combined. For example, CLI + Financial Expert Edition + Deep Review: run --agent-wizard to select finance and specify scenarios, and the code agent's own model completes a complete agent-side in-depth financial expert evaluation, with the final JSON containing domainExpert.score and verified certificates. SkillLens doesn't just check whether SKILL.md is written correctly; it evaluates a skill as a "prototype of a releasing AI product."

Overview (machine translated)

Chinese Simplified | English

Product Demonstration

https://github.com/user-attachments/assets/8bc9bfce-bfe5-4c3a-a915-9aaf1520969e

SkillLens is an open-source, self-hosted web tool used to evaluate Agent Skills. You can upload a SKILL.md, a skill folder, or a skill compressed package. SkillLens generates a quantitative report including scoring criteria, evidence, SkillLens Deep Review, market signals, and actionable improvement suggestions.

It is aimed at developers building skills for Cursor, Claude, OpenClaw, or similar Agent ecosystems, helping you answer a core question:

Is this skill truly useful, reliable, easy to find, and worth publishing?

Current Ability Map

SkillLens is composed of three quadrature dimensions and can be selected as needed:

  • What to Comment On (Review Perspective)

  • General Evaluation: 5 pillars, 100-point scale; Enable differentiated sub-dimensions by skill structure (atomic / pipeline / composite).

  • Financial Expert Edition (stackable): In addition to the general total score, it outputs domainExpert expert scores for financial scenarios, covering investment and financing, quantitative trading, stock trading, securities, banking, investment education, and more

  • How Deep (Depth of Review)

  • Rule Minute Preview: Deterministic checks (pass / partial / fail / not_applicable), returns within seconds, without consuming LLM credits

  • SkillLens Deep Review: Above the rule score, the LLM evaluates subjective dimensions; Rule items are still determined by the official scorer, preventing LLM from acting freely

  • Who Will Use It (Call Entry)

  • Web UI: For users — Browser drag-and-drop uploads to visualize reports

  • Agent CLI: Used for code agents (such as Cursor / WorkBuddy / Hermes) — --agent-wizard guides selection and prevents agents from fabricating scores through deepReviewCertificate certificate verification

The three dimensions can be freely combined. For example, CLI + Financial Expert Edition + Deep Review: run --agent-wizard to select finance and specify scenarios, and the code agent’s own model completes a complete agent-side in-depth financial expert evaluation, with the final JSON containing domainExpert.score and verified certificates.

The most worthwhile place to try

SkillLens doesn’t just check whether SKILL.md is written correctly; it evaluates a skill as a “prototype of a releasable AI product.”

This page was translated automatically. Technical details may differ from the original.

Recommended Tools

Try a different keyword or remove a filter.

Install

npx skillfish add andrewnggirl/skilllens
New to MCP? Start with the mcpskill.org guide →