Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when creating or rewriting README.md for projects. Triggers on "write README", "create README", "update README". Creates human-focused documentation with proper structure.
.claude/skills/serejaris-readme-generator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-12 | ✓→✗ | ▼ Worse | -48% | 0% |
| case-13 | ✓→✗ | ▼ Worse | 47% | 0% |
| case-14 | ✓→✗ | ▼ Worse | 37% | 0% |
| case-15 | ✓→✗ | ▼ Worse | -37% | 0% |
Generate human-focused README files following current best practices.
Core principle: README is for humans, not agents. Research → Analyze → Write.
Model recommendation: Use Task tool with model: haiku for README generation — faster and cheaper for documentation tasks.
dotdigraph readme_flow { rankdir=TB; "1. Research best practices via Exa" [shape=box]; "2. Analyze project structure" [shape=box]; "3. Identify project type" [shape=diamond]; "4. Write README with appropriate sections" [shape=box]; "5. Commit and push if requested" [shape=box]; "1. Research best practices via Exa" -> "2. Analyze project structure"; "2. Analyze project structure" -> "3. Identify project type"; "3. Identify project type" -> "4. Write README with appropriate sections"; "4. Write README with appropriate sections" -> "5. Commit and push if requested"; }
Always search first for current README best practices:
mcp__exa__web_search_exa: "best practices README.md {year} {project_type}"Look for:
Read these files to understand the project:
CLAUDE.md or existing docspackage.json / pyproject.toml / Cargo.toml| Type | Key Sections | |------|--------------| | CLI Tool | Usage examples, Options table, Installation | | Library/SDK | API overview, Quick start, Examples | | Web App | Features, Quick start, Architecture diagram | | Full-stack | Architecture, Quick start both parts, Tech stack | | AI/ML | Model info, Requirements (GPU), Examples |
Essential (all projects):
markdown# Project Name **One-line value proposition.** Brief description (2-3 sentences max). ## What It Does / Features - Bullet points of capabilities - Focus on VALUE not implementation ## Quick Start Minimal steps to get running (< 5 commands) ## Requirements - Runtime versions - System requirements - Disk space if significant ## Usage Examples for common cases ## Tech Stack (if non-trivial) Table format preferred ## License
Optional by project type:
| Section | When to Include | |---------|-----------------| | Architecture diagram | Web apps, full-stack, complex systems | | Project Structure | Open source, contribution expected | | API Reference | Libraries, SDKs (or link to docs) | | Related Skills | If Claude Code skills exist | | CLI Options table | CLI tools | | Configuration | Apps with config files | | Known Limitations | If significant gotchas exist |
DO:
DON'T:
If project has Claude Code skills, add:
markdown## Related Skills | Skill | Trigger | What it does | |-------|---------|--------------| | `skill-name` | "trigger phrase" | Brief description |
| Project has... | Include | |----------------|---------| | Web UI | Screenshot or architecture | | CLI | Options table, examples | | API | Endpoint summary (not full docs) | | AI/ML | Model, hardware requirements | | Claude skills | Related Skills table | | Complex arch | Mermaid diagram |
| Mistake | Fix | |---------|-----| | API docs dump | Link to separate file or remove | | No quick start | Add "clone → run" in < 5 steps | | Wall of text | Use bullets, tables, code blocks | | Missing requirements | Always list runtime + system reqs | | Generic description | Start with specific value prop |
markdown# Project Name **Value proposition in one line.** Brief description. ## Features - Feature 1 - Feature 2 ## Quick Start \`\`\`bash ./start.sh \`\`\` ## Requirements - OS/hardware - Runtime versions ## Tech Stack | Layer | Stack | |-------|-------| ## Architecture \`\`\`mermaid flowchart... \`\`\` ## Project Structure \`\`\` project/ ├── ... \`\`\` ## Related Skills | Skill | Trigger | Description | ## Development How to test, build ## License
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | fail→fail | 15,733 | 12,653 | -20% | 1 | 1 | 0% | 1,598 | 2,391 | +50% | 0 | 0 | — |
case-01 | fail→fail | 8,441 | 13,037 | +54% | 1 | 1 | 0% | 1,642 | 1,529 | -7% | 0 | 0 | — |
case-02 | fail→fail | 3,280 | 5,373 | +64% | 1 | 1 | 0% | 181 | 1,432 | +691% | 0 | 0 | — |
case-08 | fail→pass | 36,674 | 21,116 | -42% | 1 | 1 | 0% | 5,153 | 4,035 | -22% | 0 | 0 | — |
case-03 | fail→fail | 10,283 | 15,612 | +52% | 1 | 1 | 0% | 1,860 | 1,449 | -22% | 0 | 0 | — |
case-04 | pass→pass | 20,833 | 21,866 | +5% | 1 | 1 | 0% | 2,592 | 4,102 | +58% | 0 | 0 | — |
case-05 | fail→fail | 17,260 | 11,831 | -31% | 1 | 1 | 0% | 3,105 | 1,569 | -49% | 0 | 0 | — |
case-06 | fail→fail | 18,991 | 18,635 | -2% | 1 | 1 | 0% | 2,672 | 1,909 | -29% | 0 | 0 | — |
case-07 | pass→pass | 23,015 | 18,857 | -18% | 1 | 1 | 0% | 2,656 | 3,034 | +14% | 0 | 0 | — |
case-10 | fail→fail | 12,091 | 9,198 | -24% | 1 | 1 | 0% | 2,268 | 2,843 | +25% | 0 | 0 | — |
case-11 | fail→fail | 20,208 | 13,014 | -36% | 1 | 1 | 0% | 3,094 | 1,597 | -48% | 0 | 0 | — |
case-12 | pass→fail | 17,687 | 5,468 | -69% | 1 | 1 | 0% | 2,787 | 1,453 | -48% | 0 | 0 | — |
case-13 | pass→fail | 11,704 | 6,726 | -43% | 1 | 1 | 0% | 1,586 | 2,333 | +47% | 0 | 0 | — |
case-14 | pass→fail | 10,813 | 7,443 | -31% | 1 | 1 | 0% | 1,709 | 2,335 | +37% | 0 | 0 | — |
case-15 | pass→fail | 14,388 | 6,957 | -52% | 1 | 1 | 0% | 2,693 | 1,685 | -37% | 0 | 0 | — |
case-16 | pass→fail | 16,768 | 9,620 | -43% | 1 | 1 | 0% | 2,393 | 1,612 | -33% | 0 | 0 | — |
case-17 | pass→fail | 11,859 | 6,955 | -41% | 1 | 1 | 0% | 2,668 | 1,404 | -47% | 0 | 0 | — |
case-18 | fail→fail | 14,042 | 7,739 | -45% | 1 | 1 | 0% | 2,527 | 1,428 | -43% | 0 | 0 | — |
case-19 | pass→pass | 20,962 | 12,257 | -42% | 1 | 1 | 0% | 3,732 | 3,216 | -14% | 0 | 0 | — |
case-20 | pass→fail | 9,383 | 6,227 | -34% | 1 | 1 | 0% | 1,978 | 1,482 | -25% | 0 | 0 | — |
case-21 | pass→fail | 13,513 | 8,107 | -40% | 1 | 1 | 0% | 2,523 | 1,502 | -40% | 0 | 0 | — |
case-22 | pass→pass | 10,540 | 11,939 | +13% | 1 | 1 | 0% | 1,414 | 2,541 | +80% | 0 | 0 | — |
case-23 | pass→pass | 4,900 | 4,994 | +2% | 1 | 1 | 0% | 897 | 1,813 | +102% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 10 counted toward the lift figure. The other 13 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of -30 percentage points is the difference between those two pass rates over the 10 comparable cases. 9 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.