SkillForge skill
A skill for creating skills that proves them by behavior: before-and-after runs, regression evals and ecosystem audit
- 918 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Runs tasks and helper scripts
- Can install optional hooks into the agent's settings
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/skillforgeDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Runs tasks and helper scripts
- Can install optional hooks into the agent's settings
This is third-party code. Review the repository files before installing.
What it does
SkillForge creates and improves skills on the principle that a skill's quality is agent behavior, not document text. The pipeline first runs a task without the skill as a baseline, then with it, and the behavioral difference becomes the gate for done. Formal checks live in a linter, and one adversarial reviewer looks for what the linter misses. The skill also routes any skill-related request: use, improve, create, compose; it watches the health of your whole collection, compiles skills for different runtimes and can act as an opt-in, unobtrusive advisor. It installs into the Claude Code or Codex skills folder and needs only Python from the standard library.
Who it is for. For developers and skill authors who want a way to create and verify skills by real effect.
Good fit when
- You need to create a new skill and confirm it helps
- You need to improve or compose several skills
- You need to check the health of your skill collection
Not a fit when
- You need a single ready skill, not a tool to create them
- You cannot run test executions
Example request
Create a skill for reviewing commits and show the agent does better with it than withoutLimitations
Evaluation is built on task runs, so it needs examples and agent executions, which use tokens. The advisor and hooks are optional and enabled manually; the skill is not auto-invoked. It needs Claude Code or the Codex CLI and Python 3.8.
How to disable. Remove the ~/.claude/skills/skillforge folder and the ~/.claude/commands/skillforge.md command file.
FAQ
How is this different from plain skill generation?
SkillForge verifies a skill by running it: it compares agent behavior without and with the skill, and that difference decides if it is done.
Is the skill auto-invoked?
No, it routes on request, and the advisor and hooks are enabled manually.
Discussion
No questions or reviews yet.
Related
SkillsA skill that creates and reviews other skills, weighing usefulness, trigger boundaries, setup and stability before release.
A skill that helps an agent author other skills to the agentskills.io spec: structure, metadata, progressive disclosure and validation
A skill that designs, tests, evaluates and packages other agent skills, starting from architecture
A meta-skill that takes a vague idea to a finished agent skill: it designs, writes the files, checks them and refines by evidence
A working folder for Claude Code where the agent asks questions and writes a new skill by Anthropic's guidance, not from memory.
A plugin of seven skills for Claude Code and Codex: prompts, subagents, skill authoring, editing, web search and a browser.
README in short
The README describes SkillForge as a skill creator that proves its skills work. Version 6 is a rework after an external audit: skills are now executed, a baseline run and a with-skill run give a behavioral delta as the gate, formal checks live in a linter, and an adversarial reviewer catches the rest. The skill routes skill-related requests, watches collection health and compiles across runtimes. Install is into the Claude Code or Codex skills folder. MIT licensed.