Launch your GitHub repo like a product — with proof, not hype.
Live self-audit · Launch evidence · Discussion · Agent Skill
LaunchSmith is a local, evidence-first launch-readiness auditor for open-source repositories. It inspects real repository files and GitHub metadata, scores the gaps that block adoption, then generates a reviewable launch pack without inventing claims.
Most repository launch advice stops at “improve the README.” LaunchSmith checks the full path from discovery to trust:
- Can a visitor understand the outcome in ten seconds?
- Is there a copy-paste route to the first useful result?
- Does the repository show real screenshots, tests, releases, and limitations?
- Are license, security, contribution, and issue paths ready?
- Can every launch claim be traced to a file, command, screenshot, or GitHub URL?
The result is a prioritized forge report instead of a generic checklist.
When GITHUB_STEP_SUMMARY is set (GitHub Actions), gh launchsmith audit appends a compact Markdown summary — score, category table, and unresolved findings — with no extra flags or runtime dependencies.
- uses: actions/checkout@v4
- run: gh extension install Lwu0rup/gh-launchsmith
- run: gh launchsmith audit --no-githubRequirements: GitHub CLI and Node.js 18 or newer.
gh extension install Lwu0rup/gh-launchsmithUpgrade later with:
gh extension upgrade launchsmithgit clone https://github.com/Lwu0rup/gh-launchsmith.git
cd gh-launchsmith
node ./gh-launchsmith --helpRun this inside any Git repository:
gh launchsmith audit --html .launchsmith/report.htmlExpected output:
LAUNCHSMITH
owner/project 76/100 NEARLY FORGED
Positioning ███████████████░░░ 82%
Proof ████████████░░░░░░ 67%
Onboarding ████████████████░░ 88%
Trust █████████████░░░░░ 72%
Distribution ██████████░░░░░░░░ 58%
Priority fixes (6)
FAIL README shows the product
Add a real screenshot, GIF, terminal capture, or short video near the top.
Then generate the reviewable launch pack:
gh launchsmith packLaunchSmith writes .launchsmith/ with:
audit.json Full machine-readable audit
report.html Self-contained visual forge report
launch.json Explicit facts, claims, and verification slots
LAUNCH_CHECKLIST.md Priority-ordered release gate
EVIDENCE.md Claim-to-evidence register
ANNOUNCEMENT.md General launch draft
HACKER_NEWS.md Show HN draft
REDDIT.md Community-first draft
X.md Short launch draft
RELEASE_NOTES.md First-release draft
Generated placeholders stay visibly marked as [ADD: ...]. LaunchSmith will not turn missing evidence into confident marketing copy.
| Command | Result |
|---|---|
gh launchsmith audit |
Terminal scorecard and prioritized fixes |
gh launchsmith audit --json |
Complete JSON audit on stdout |
gh launchsmith audit --html report.html |
Offline interactive-style HTML report |
gh launchsmith pack |
Evidence register, launch drafts, checklist, report |
gh launchsmith preview |
Regenerate .launchsmith/report.html |
gh launchsmith watch |
Append a local GitHub metrics snapshot |
Useful options:
--cwd <path> Inspect another repository
--no-github Run with local evidence only
--strict Exit non-zero below 80/100
--force Replace generated launch-pack files
--output <path> Choose the pack/report/metrics location
| Area | Examples of inspectable evidence |
|---|---|
| Positioning | README opening, GitHub description, audience, use cases, topics |
| Proof | Demo assets, runnable commands, tests, CI, expected output, releases |
| Onboarding | Installation, quick start, examples, prerequisites, troubleshooting |
| Trust | License, security policy, contribution guide, conduct, change history |
| Distribution | Social preview, homepage/demo, issue templates, launch materials |
Each rule returns its score, the evidence LaunchSmith observed, and one concrete next action.
The repository includes skills/launchsmith, compatible with agent-skill workflows for Codex and Claude Code.
npx skills add Lwu0rup/gh-launchsmithExample prompt:
Use $launchsmith to audit this repository, fix the highest-value launch gaps,
capture real proof, and prepare a reviewed launch pack.
The skill runs the deterministic audit first, improves the repository from evidence, and reruns the audit before preparing public copy.
LaunchSmith deliberately separates facts from creative work:
- Local files and Git metadata are collected.
- GitHub metadata is queried through your authenticated
ghsession. - Deterministic rules produce findings and scores.
- Launch materials cite observed evidence or retain an explicit placeholder.
The default audit is read-only. LaunchSmith does not run arbitrary repository commands, publish releases, post to communities, buy engagement, or automate stars.
Change into the project root, or pass --cwd /path/to/repo.
Authenticate the GitHub CLI and ensure the repository has a GitHub origin remote:
gh auth login
git remote -vLocal auditing still works with --no-github.
Review the existing files first. Use --force only when replacement is intentional.
npm test
node ./gh-launchsmith audit --no-github
node ./gh-launchsmith pack --no-github --output /tmp/launchsmith-packLaunchSmith has zero runtime dependencies and uses the Node.js standard library.
- GitHub Actions job-summary output
- Verification commands with explicit allowlists and saved receipts
- Windows GitHub CLI extension smoke test
- Repository-specific demo capture recipes
- Compare reports across releases
- Additional agent adapters and launch-pack templates
See CONTRIBUTING.md. Bug reports should include the command, platform, Node.js version, and a redacted audit fixture when possible.