Snow7-G

SkillSeam — Development skill for Claude Code

Development community

Simulate how your agent picks skills and see which skill steals whose tasks, before your users do.

How to install SkillSeam

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Snow7-G/SkillSeam and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What SkillSeam does

Simulate how your agent picks skills and see which skill steals whose tasks, before your users do. Works with Claude Code, Codex and any SKILL.md-based runtime.

Alternatives in Development

Understand Anything — Graphs that teach graphs that impress 80.7k ★
Ralph Loop — Autonomous coding loop — works through tasks, commits, repeats 14k ★
Claude SEO — Multi-Platform Agent Instructions — For Cursor, Cursor Cloud Agents, Google Antigravity, and Gemini CLI 5k ★

README

English · 简体中文 · 日本語 · 한국어 · Español · Deutsch

Simulate the way your agent picks skills, and find out which skill steals whose tasks. Before your users find out.

The problem

Agent runtimes like Claude Code and Codex load skills by reading one or two lines of description. When a task comes in, the agent picks whichever skill sounds closest.

That works fine until two descriptions overlap. In our hospital demo, a booking skill's description claimed it also handled "vision report inquiries". The report-reading skill owned that job. So a patient asks for a report reading, the booking skill answers, and then it makes things up. Nothing errors, and nothing gets logged. We only found out because patients complained.

SkillSeam replays that selection process before you ship. On a 40-task demo with 6 skills, it exposed 4 planted conflicts at 5/5 vote consistency, and every clean task passed. Each conflict report points at the exact keywords in the thief's description that caused the theft.