Decide if it's even worth it, then introduce your private skill to the world — found, understood, verified, used, and improved upon.
world-intro is a gated, end-to-end pipeline that turns "I use it and it's great" into "strangers want it." It doesn't just write a README — it tells you whether to open-source at all, generalizes your tool out of your own machine, proves it on real data that becomes the case studies, polishes the narrative against anti-AI-slop rules, ships it, and promotes it so it actually gets seen.
Original plan: push it to GitHub and hope. Which is how most tools get 0 stars next to an existing project that already does it better. What it does: P0 searches the real field first. It returns one of three verdicts — empty lane (ship fast), crowded-but-unowned (ship, but name your difference in the README), or already-owned (don't — contribute upstream instead). You build only when the search says build.
Original plan: copy the folder to a public repo and spend a weekend deleting things until it stops leaking secrets. What it does: P2 strips the personal matrix — inlines private libs into a single zero-dependency file, turns your auth into optional standard env vars, makes paid sources optional with a zero-config default path — and leaves commented-open spots that invite PRs.
Original plan: tweet the link once and refresh the star count. What it does: P6 narrows the value prop to one sentence, writes platform-fit copy through a de-slop pass, and produces real evidence images (HTML-rendered so repo names and numbers don't garble) plus a publishing manifest for each platform.
These aren't hypotheticals — they're the two projects this pipeline actually shipped.
# 1. Install (Claude Code shown; for other agents, add SKILL.md to the system prompt)
git clone https://github.com/a28939876-max/world-intro
cp -r world-intro ~/.claude/skills/world-intro2. Tell your agent:
"Open-source this skill for me: <path-or-description of your tool>"
3. Answer the three gate questions (name / language / which real object to prove it on),
and step through the verdict → generalize → prove → polish → publish → promote pipeline.
| Without world-intro | With world-intro | |
|---|---|---|
| Should you ship? | A hunch | A three-way verdict from a real competitor search |
| Getting it public | Delete secrets by hand until it stops leaking | Matrix-stripped to a zero-config single file |
| Credibility | "Trust me, it works" | README cases are real runs with real numbers |
| The README | A feature list | Persuasion-first: need → result, plain words, weak spots in the FAQ |
| After the push | 0 stars, a single tweet | Narrowed pitch + per-platform copy & images + a manifest |
| Gates | None — you find out it leaked / mis-shipped later | Three human gates: plan / publish / promote |
world-intro didn't appear from nowhere. It's the distilled pipeline behind three real releases — skill-lineage, world-aid and skill-hunter-company — and every rule here was paid for in those launches. It was then pointed at itself: the verdict you're reading, the case studies below, this very README, were produced by running world-intro on world-intro.
flowchart LR
P0["P0 · worth-it search<br/>(ship / differentiate / don't)"] --> G1{gate ①<br/>name·lang·scope}
G1 --> P2["P2 · generalize<br/>strip the matrix"]
P2 --> P3["P3 · real acceptance<br/>run → backfill cases"]
P3 --> P4["P4 · polish<br/>narrative + challenge FAQ"]
P4 --> G2{gate ②<br/>account·repo}
G2 --> P5["P5 · publish"]
P5 --> P6["P6 · promote<br/>copy + images + manifest"]
P6 --> G3{gate ③<br/>per-platform}
The skill itself is SKILL.md; the depth lives in pipeline/ — polish rules, promotion playbook, publish pitfalls.
Typical picks from the launches world-intro has powered — full write-ups in cases/:
| Case | The tool | A real result from its acceptance run |
|---|---|---|
| 01 · skill-lineage | trace a skill's forks/mirrors/injections before you install | found a 5,233★ localized fork that star-sorting hides; one diff caught a "silently POST a score back" injection |
| 02 · world-aid | say a need, it rounds up the existing tools | "YouTube → text": 13 candidates → 9 distinct; the flashiest 7★ "Tor" pick was caught doing sudo systemctl start tor |
| 03 · skill-hunter-company | the whole find→vet→fuse→install→govern lifecycle, as a "headhunting firm" for skills | a 4-round model-vs-model debate overturned a "don't open-source the orchestration" verdict; the umbrella shipped orchestrating the two repos above |
Skills like open-source-hardening-skills make your code clean — tests, CI, refactor, governance docs. world-intro handles what comes before and after that: whether it's worth open-sourcing, how to generalize it out of your environment, how to turn real results into a narrative people want, and how to get it seen once it's live. They're upstream/downstream complements — run both.
"This is a methodology skill, not a binary tool — where's the real value?" The value is the rules that were paid for in real launches: the matrix-stripping table, the anti-slop checklist, "evidence images must be HTML-rendered, not text-to-image," the three gates. It's opinionated on purpose — most OSS-launch advice is human-read prose that assumes you've already decided to ship; this one searches first and will tell you not to.
"Won't the official/leading project just absorb my niche?" If P0 says the lane is owned, that's the correct outcome — world-intro tells you to contribute upstream instead of shipping a redundant repo. When it says ship, it makes you name the difference in the README so absorption is a fair fight, not a surprise.
"Are these really only three cases?" No — these three are typical picks from the launches the pipeline has run; the rules generalize to any private skill or internal tool, not just these.
MIT — see LICENSE.