Agent Skill Inspector

Drop a skill zip and get a full report: score, spec check, security scan, token cost and compatibility.

Private by design

All analysis runs in your browser, on your own device. Your files never leave the device and are never uploaded to any server. Cubbbix never sees them.

Drop a skill .zip here

Or pick a zip, or pick the loose skill files from a folder. No account, no upload.

Choose a .zip file Choose skill files

What gets checked

Agent Skill Score

A score from 0 to 100 with a visible breakdown, so you can see what to fix first.

Spec and structure

SKILL.md at the skill root, valid frontmatter, and the usual scripts, references and assets folders.

Security scan

Risky patterns with severity and exact file and line, plus a count for each category.

Context cost

An estimated token count for the whole skill and per file, with the largest files highlighted.

Portability

Hardcoded paths, OS specific commands and platform specific fields that tie the skill to one machine.

Compatibility

Claude, Codex, Copilot and Cursor, each shown as supported, partial or unknown, with a reason.

A note on trust: this page reads text patterns. It cannot prove a skill is safe, and it cannot run the skill. Always read flagged lines yourself before you run any script from a skill you did not write.

Frequently Asked Questions

An agent skill is a small folder of instructions that teaches an AI agent how to do one job well. It usually starts with a SKILL.md file that names the skill, describes when to use it, and points to extra scripts, reference notes and assets.

The score runs from 0 to 100 and combines four parts: spec compliance (40 points), structure (20 points), portability (20 points) and security (20 points). The breakdown is shown under the score, so you can see exactly what pulled it up or down.

The scan looks for risky patterns, such as destructive commands, credential access, remote downloads and prompt injection wording, and it shows the exact file and line for each hit. It is a pattern check, not a guarantee. A clean result does not prove a skill is safe, and a hit needs a human look before you run anything.

Token cost is an estimate only. It divides the character count of each file by 4, which is a rough rule of thumb for English text. Real token counts vary by model and by content, so treat the number as a guide for context cost, not an exact bill.

No. All analysis runs in your browser on your own device. The zip is opened with JavaScript, nothing is uploaded to any server, and Cubbbix never sees your files.

Claude, Codex, Copilot and Cursor. Each one is shown as supported, partial or unknown, with a reason. Only Claude fully reads the standard frontmatter fields, so other platforms often need a test run before you rely on the skill there.

Related Tools

AI Tools

Browse the free AI utilities on Cubbbix.

Word Counter

Count words and characters in any text instantly.

All Tools

See the full catalog of free Cubbbix tools.

Was this tool helpful?

Comments

Loading comments...

Check Out Other Popular Tools

Bra Size Chart

Calculate your perfect bra size instantly and convert between US, UK, and EU measurements. Interactive sizing charts included.

Health & Fitness93

ASCII Table & Reference

A complete reference of ASCII characters, including control codes and extended ASCII with Dec, Hex, and Binary values.

Developer Tools143

スタバメニュー価格・カロリー・カスタム計算機

スターバックス（スタバ）の最新メニュー一覧、値段（持ち帰り8%・店内10%）、サイズ別カロリー、おすすめカスタムを検索・計算。フラペチーノ・ラテ・ティーの料金シミュレーター。

Menus204