skill-builder
Skill0 starsChanged 3 months agoMITView on GitHub
Turns a process the operator keeps repeating by hand into a real skill instead of one more one-off prompt they'll have to re-explain next time. Fires on requests like "my agent doesn't know how to do this yet", "I keep typing the same instructions every time", "turn this into a skill", "can you make a skill for this", "I do this every week and it's always slightly different", "build me a command for X", "why do I have to explain this from scratch every time", "make this repeatable", or "I wrote this out again, can we automate it". Also fires when the operator describes a multi-step task they say they do "every time" or "constantly" without naming a skill that already covers it.
73 other projects use a version of this.
Adopted by
Most recently taken up by
- arctan303/creator.skilltheir own version, first seen Nov 12, 2026
- Kikobazz123/lordgen-skill-buildertheir own version, first seen Nov 8, 2026
- AsiaOstrich/EngramGraphtheir own version, first seen Nov 8, 2026
- shetengteng/skillix-hubtheir own version, first seen Nov 8, 2026
- davidgfolch/AI-job-searchtheir own version, first seen Nov 8, 2026
- RKelln/hermes_skillstheir own version, first seen Nov 7, 2026
Used with
Projects with it also use contract-test, literature-research and 9 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
squall-chuaSkillcontract-test
Test a running API against its own contract — OpenAPI, GraphQL, protobuf, or AsyncAPI — from the outside, and report every drift between what the contract promises and what the API does.
Also in 3 projects that use this
nickjwellsSkillliterature-research
Search academic literature and perform deep research reviews. Use when user asks to search PubMed, find academic papers, or do literature reviews.
Also in 3 projects that use this
nopandererSkillcommit-and-push
Automatically analyze git changes, generate Conventional Commits compliant commit messages, commit changes, and push to remote in a single automated workflow. Use when the user invokes /commit-and-push or explicitly requests to automatically commit and push changes following Conventional Commits format (feat:, fix:, docs:, etc.). This skill provides a fully automated git workflow without manual intervention.
Also in 3 projects that use this
MicrosoftSkillskill-reviewer
Review skill PRs with structured severity-rated feedback covering token budgets, routing conflicts, and repo conventions. WHEN: \"review skill\", \"review skill PR\", \"review skill changes\", \"check skill quality\", \"skill PR feedback\".
Also in 3 projects that use this
SkillPilotCHSkilllevel-up
Eine Automatisierung pro Woche finden, scopen, shippen. Output landet als proposed-Eintrag in cadence.md mit konkreten Implementations-Schritten.
Also in 3 projects that use this
MicrosoftSkillexcalidraw-diagram
Create Excalidraw diagram JSON files that make visual arguments. Use when the user wants to visualize workflows, architectures, or concepts.
Also in 4 projects that use this
Other versions73 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 73
- yusufkaraaslan/Skill_Seekers★ 15ksame name
- block/berd★ 1.1ksame name
- boshu2/agentops★ 454same name
- metaskills/skill-builder★ 112same name
- AsiaOstrich/universal-dev-standards★ 84same name
- nateherkai/a-bunch-of-skills★ 41same name
- clasen/Skills★ 40same name
- sanqi-cd/Sanqi-Skills★ 28same name
- arctan303/creator.skill★ 19same name
- aiagentwithdhruv/skills★ 16same name
63 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
12 rules. 0 are common, 12 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 64 edited versions we read keep these.
12
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 18
Confirm this is one process, not two: Ask what the operator wants automated, in their own words.
- Distinctive: line 24
Do not collapse blocks into one wall of questions and do not skip a block because the answer seems obvious — an obvious-seeming answer is …
- Distinctive: line 26
Wait for the operator's reply before sending the next block.
- Distinctive: line 28
Block 1 — Identity: Ask for the skill's name (the directory name it will live at) and a one-line purpose.
- Distinctive: line 31
Block 2 — Persona: Ask who the skill is being when it runs, and what it's expert in.
- Distinctive: line 34
Block 3 — Task: Ask for the single job in one sentence.
- Distinctive: line 36
Block 4 — Inputs: Ask what the operator will hand the skill each time, and — separately — what the skill should do when that input is …
- Distinctive: line 40
Block 5 — Instructions: Ask for the ordered steps the operator actually follows today, by hand.
- Distinctive: line 43
Block 6 — Output: Ask for the exact output format and the exact destination path.
- Distinctive: line 46
Block 7 — Tools: Ask what the skill needs to read, what it needs to write, and what external tool or service (if any) it needs to call.
- Distinctive: line 57
Block 9 — Validation: Ask how the operator will know a run was good, and ask for one concrete example of a bad output to avoid.
- Distinctive: line 84
Keep the whole file under 200 lines.
“Common” is measured: kept by at least 6 in 10 of the 64 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.
Files
Skill Builder
The failure this prevents has two shapes, not one. Either the skill you write never fires, because its description doesn't say what the operator actually types when they hit this problem — or it fires and produces the wrong thing, because the task was ambiguous and the model filled the gaps with its own guess instead of the operator's intent. Both failures are cheap to prevent and expensive to live with: a clarifying question costs the operator twenty seconds, once. A skill that guesses wrong costs those twenty seconds back, plus a rewrite, every single time it runs after that. This skill exists to ask before it writes.
Procedure
- Confirm this is one process, not two. Ask what the operator wants automated, in their own words. If the answer describes two jobs joined by "and" — draft the thing and send the thing, research and write — stop and say so: that's two skills, not one. Get agreement on the single job before block 1.
Run the interview as nine separate messages, one block per message, in this exact order. Do not collapse blocks into one wall of questions and do not skip a block because the answer seems obvious — an obvious-seeming answer is exactly the kind that turns out to be wrong. Wait for the operator's reply before sending the next block.
- Block 1 — Identity. Ask for the skill's name (the directory name it will live at) and a one-line purpose. Reject a name that's a verb-less noun ("reports") — push for a name that names the action ("write-status-report").
- Block 2 — Persona. Ask who the skill is being when it runs, and what it's expert in. A skill with no persona defaults to a generic assistant voice, which is how client-facing output ends up sounding like nobody in particular.
- Block 3 — Task. Ask for the single job in one sentence. If a second sentence shows up with "and", that's the two-skills problem from step 1 again — resolve it before moving on.
- Block 4 — Inputs. Ask what the operator will hand the skill each time, and — separately — what the skill should do when that input is missing or incomplete. "Ask the operator" is a valid answer here; "guess" is not, and if the operator says "just figure it out," push back once before accepting it.
- Block 5 — Instructions. Ask for the ordered steps the operator actually follows today, by hand. Write these down as the numbered procedure verbatim — don't paraphrase into something more generic. A step the operator didn't actually describe is a step you invented.
- Block 6 — Output. Ask for the exact output format and the exact destination path. Not "a summary" — what kind of file, in what shape, saved where. If the answer names a folder this repo doesn't have yet, flag it now, not at step 5 below.
- Block 7 — Tools. Ask what the skill needs to read, what it needs to write, and what external tool or service (if any) it needs to call. Write the shortest tool list that covers what was described — an unused tool listed "just in case" is a tool the model will try to justify using.
- Block 8 — Configuration. Collect at least five distinct trigger phrases, in the operator's own words — not your paraphrase of their intent. This is the block where firing is won or lost: a description full of correct-sounding generic language is the single most common reason a skill never fires. Explicitly ask how the operator would complain about this problem when it's bugging them ("I keep having to…", "why do I always…"), not only how they'd request the solution — a complaint and a command read completely differently to a description match.
- Block 9 — Validation. Ask how the operator will know a run was good, and ask for one concrete example of a bad output to avoid. A bad-output example is worth more than another paragraph of instructions, because it rules out a failure mode instead of describing a success.
- Check the target folder against the routing map before writing. Read
CLAUDE.md's routing map. If the destination path from block 6 lands inside a folder the map already lists, use it as-is. If it doesn't — a new top-level folder, or a new subfolder the map is silent on — tell the operator you're adding a row for it, and do so in the same pass (see step 5). Writing to an unmapped folder without updating the map recreates the exact failure/route-fixexists to catch: a file nobody can find because nothing points at it. - Write
.claude/skills/<name>/SKILL.mdusing the nine answers directly — every block maps to a part of the file:- Frontmatter
namefrom block 1;descriptionbuilt from block 8's five-plus phrases plus one line of context from block 3, written the way a user types, not the way a spec reads. - Opening: one or two sentences naming the failure the skill prevents (drawn from what goes wrong today without it — usually visible in blocks 3 and 9's bad-output example), not a restatement of the task.
- Numbered procedure from block 5, each step naming what to read, what to ask, what to write — no step that says "handle it appropriately" or "as needed."
- Output section from block 6 with the literal path.
- Tools noted inline wherever a step uses one, matching block 7 — don't list a tool the procedure never calls.
- A
## When NOT to run thissection naming at least one skill from this kit's seven that the new skill could get confused with, and the one-line distinction — ask the operator if it isn't obvious from the interview. - Keep the whole file under 200 lines. If the interview produced more than that, the task was probably still two jobs — go back to step 1.
- Frontmatter
- Add the routing row, if step 3 found a new folder. Edit
CLAUDE.md's routing map table in the same pass: new folder name, what it holds, when to go there. Do this before showing the finished skill, not after — a skill that writes to a place the map doesn't mention is invisible the moment this session ends. - Show the finished file to the operator before treating this as done. Print the full
SKILL.md, not a summary of it. Ask them to try the exact trigger phrases from block 8 against the description out loud — if any of the five don't obviously match, fix the description now, because this is the cheapest moment this skill will ever have to fix it.
When NOT to run this
/leverage is for finding what's worth automating in the first place — it surfaces friction
across a week of work and decides what's worth building. This skill is for after that
decision is made: the operator already knows the process, already knows it repeats, and needs
it turned into a real skill. If the operator isn't sure yet what to automate, send them to
/leverage first. This skill also never runs an existing skill on the operator's behalf — if
a skill covering this already exists on disk, say so and stop instead of writing a duplicate.
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
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.

