The problem: You've built something useful. Self-publishing rarely works. Community review is free QA — if you know where to submit.
The solution: Extract → Find gap → PR hardening → Launch. 4 rounds of bot review found 9 bugs in a 200-line script.
Note: This is a methodology document, not an executable tool. The steps below are a workflow guide — follow them with your own code.
Any "no" → stop. Ask someone else — self-assessment of originality is biased.
- Original? Not a wrapper. Credit ≠ originality. If you wouldn't present it as your own in an interview, don't submit.
- General? No personal paths, secrets, or machine-specific config.
- <500 lines? Smaller = faster review. If larger, extract the core.
Strip personal data. Stdlib preferred. 3-line usage at top. Gate: Stranger runs it cold. No secrets. Clear enough for 6 months from now.
Complementary gap, not redundant. Search:
github.com/search: "claude code" stars:>1000 path:skills
Evaluate: Stars >1K, commits <30d, has review bots, gap confirmed, external PRs accepted.
/.github/workflows/ and closed PRs for bot activity.
Skip if: >90 days stale, 50+ open PRs, 3+ feature overlaps.
Match their format exactly. Single commit. Respect maintainer time — small PRs, prompt responses to feedback.
| Review source | Catch | Response |
|---|---|---|
| Bots | Structural bugs, edge cases, compat | Minutes |
| Human | Architecture, naming, scope | Days-weeks |
| Community | Real-world patterns | Ongoing |
No bots? Named-persona review (see Pre-Flight). Stuck 2+ weeks? Comment. 4 weeks? Launch with review alone. Rejected? Fix, resubmit — feedback is valuable.
Push hardened code. README: what/why/install/examples. License compatible. Link PR. Attribute borrowed patterns.
After launching: What worked? What broke? Update validated cases. The flywheel improves with every cycle.
- Code deps: Link upstream, state what it provides
- Format patterns: "Inspired by [repo]'s [pattern]" in README
- Reviewers: Thank by name in commit messages
- This method borrows: agent-skills invented Rationalizations + Red Flags format — credited
✅ delivery-gate: ECC (100K★, ✅bots) → 4 rounds, 9 bugs → repo
✅ self-audit: anthropics/skills (154K★,
Q: PR rejected? Read reason. Fix. Resubmit. Q: No community fits? Too niche, or search broader — try adjacent ecosystems. Q: Why not self-publish directly? Bots catch bugs you'll never find alone. Our 200-line script had 9 bugs — 8 found by bots.
gategrow/checkgrow — quality framework hub · imprint — identity persistence · delivery-gate · dual-pool-review · opensource.guide