juanmarchetto/agent-skills/oss-readiness

OSS Readiness — Open Source Readiness Evaluator

Evaluate whether your project is ready to be open-sourced. Five specialist agents audit your codebase simultaneously, then a synthesizer produces a unified readiness score with prioritized fixes sorted by severity.

How It Works

Input is a project path. Five agents scan the project in parallel:

1. License Auditor (agents/license-auditor.md) — checks LICENSE file presence and type, dependency license compatibility (copyleft contamination in permissive projects), attribution requirements, patent clauses, files with conflicting license headers
2. Documentation Reviewer (agents/docs-reviewer.md) — README completeness (description, install, usage, examples, badges), CONTRIBUTING.md quality, CHANGELOG.md presence, API documentation for libraries, code comment quality on public functions, quick-start guide
3. Security Scanner (agents/security-scanner.md) — hardcoded secrets (API keys, tokens, passwords) in code, .env files not in .gitignore, secrets accidentally committed in git history, known vulnerable dependencies, debug/test credentials left in code, overly permissive file permissions
4. Community Readiness (agents/community-readiness.md) — issue templates, PR templates, CODE_OF_CONDUCT.md, governance model, onboarding friction (steps from clone to running), labels/milestones setup, first-time contributor friendliness
5. API Surface Analyzer (agents/api-surface-analyzer.md) — public API clearly defined (exports, public modules), semver compliance readiness, breaking change risk, backwards compatibility, type definitions/interfaces documented, deprecation policy

After all five agents complete, the Synthesizer (agents/synthesizer.md) combines their outputs into a unified assessment with weighted scoring, blocker identification, and a phased release plan if the project is not ready.

Usage

"Is my project ready to open source? ./my-project"
"Evaluate open source readiness for this repo"
"What do I need to fix before publishing this code?"
"Prepare ./my-library for public release"
"Open source checklist for this project"

Output Format

The evaluation produces a structured report:

OPEN SOURCE READINESS EVALUATION
Project: [name]
 
Overall Readiness: 7.2/10 — CONDITIONAL GO
 
| Dimension      | Score | Critical Issues                          |
|----------------|-------|------------------------------------------|
| Licensing      | 9/10  | MIT license, all deps compatible         |
| Documentation  | 6/10  | README good, missing CONTRIBUTING.md     |
| Security       | 4/10  | 2 hardcoded API keys found in git history|
| Community      | 5/10  | No issue templates, no CoC              |
| API Surface    | 8/10  | Stable public API, semver ready          |
 
BLOCKERS (fix before publishing):
1. [SECURITY] Remove API keys from git history (use BFG Repo Cleaner)
2. [SECURITY] Add .env to .gitignore
 
HIGH PRIORITY:
3. [DOCS] Create CONTRIBUTING.md
4. [COMMUNITY] Add issue templates
5. [COMMUNITY] Add CODE_OF_CONDUCT.md
 
RECOMMENDED:
6. [DOCS] Add API documentation
7. [COMMUNITY] Define governance model

This is followed by:

• Blocker details with specific fix instructions
• Phased release plan if the project is not ready
• Quick wins that can be fixed in under 30 minutes
• Verdict: GO / CONDITIONAL GO / NOT READY

Agent Files

References

Error Handling

• No LICENSE found: License Auditor flags this as a BLOCKER and recommends license selection based on project type and dependency licenses
• No README found: Documentation Reviewer flags as a BLOCKER and provides a README template tailored to the project
• No git history: Security Scanner skips history analysis but warns that uninitialized repos cannot be verified for leaked secrets; other checks proceed normally
• No package manager: License Auditor skips dependency license scan but notes it as a limitation; manual dependency review is recommended
• Binary/non-code project: API Surface Analyzer returns "Not Applicable" and the synthesizer redistributes its weight across the other four dimensions

When a dimension cannot be fully assessed, the agent returns a partial score with clear notation of what was checked vs. what was skipped. The synthesizer adjusts weights accordingly.

Scoring Weights

The synthesizer uses weighted scoring reflecting real-world open-source risk:

Verdict Thresholds

Language

All agents respond in the same language the user writes in. Detected automatically — no configuration needed.