agentleFS
bonez-iobonez-io/re_gent/.claude/skills/skill-factory/SKILL.md

skill-factory

Skill4 starsChanged 6 months agoApache-2.0View on GitHub

Create a new re_gent skill from a plain-language description, writing a valid SKILL.md. Use when the user wants a new skill, or to build one that style-factory proposed.

14 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 13 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before writing, be able to answer in one sentence each:

    line 15
  2. Distinctive:

    If the request is vague, ask one question.

    line 25
  3. Very specific:

    Use Glob to list .claude/skills/*/SKILL.md (or the active host's matching skills directory), then read any neighbour that looks close.

    line 30
  4. Very specific:

    Prefer real commands you have verified exist: Check with rgt <cmd> --help before referencing a flag.

    line 57

“Common” is measured: kept by at least 6 in 10 of the 13 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Turn a description into a working skill file.

A skill is a prompt plus a tool grant. The intelligence is the agent's, the data is re_gent's, and the skill is the wiring between them — so writing one is mostly deciding what to ask for and what it is allowed to run.

1. Settle the shape

Before writing, be able to answer in one sentence each:

  • Trigger — when should an agent reach for this? This becomes description, and it is the single most important line in the file: it is what the agent matches against. Name the situation, not the mechanism.
  • Input — a path, a step hash, a symptom, nothing? This becomes argument-hint.
  • Data — which rgt commands supply the facts.
  • Output — what the user should be holding at the end.

If the request is vague, ask one question. A skill built on a guess triggers at the wrong moment and is worse than no skill.

2. Check it is not already there

Use Glob to list .claude/skills/*/SKILL.md (or the active host's matching skills directory), then read any neighbour that looks close.

Extending an existing skill beats adding a near-duplicate — two skills with overlapping descriptions make the agent pick badly.

3. Write it

.claude/skills/<kebab-case-name>/SKILL.md:

---
description: <what it does + when to use it, one sentence>
allowed-tools: Bash(rgt <command> *)
argument-hint: "<arg>"
---

<Instructions to the agent.>

Rules that matter:

  • Grant the narrowest tools that work. Bash(rgt log *) — never bare Bash. A skill that only reads should never be able to write.
  • Write instructions, not prose. The reader is an agent about to act.
  • Say what to do with the data, not just how to fetch it. The gathering is the easy half; the analysis is the skill.
  • Prefer real commands you have verified exist. Check with rgt <cmd> --help before referencing a flag. A skill that cites a flag that does not exist fails at the moment it is trusted.
  • Name sibling skills the agent should chain to, so skills compose.

4. Verify before claiming it works

rgt <each command in allowed-tools> --help

Confirm every command and flag exists. Then state plainly what you could not test — a skill's real behaviour only shows when an agent runs it against real history.

Report

Path written, the description line verbatim (so the user can judge when it will trigger), the commands it may run, and one example invocation.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.