writing-project-docs
A documentation tool that creates short, high-impact summaries for features or PRs to ensure quick team understanding.
Install
mkdir -p .claude/skills/writing-project-docs && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/13908" && unzip -o skill.zip -d .claude/skills/writing-project-docs && rm skill.zipInstalls to .claude/skills/writing-project-docs
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Use when generating documentation for a feature, PR, or implementation — especially when docs are too long, hard to read, or need to be understood by non-technical reviewers, testers, or team leadsKey capabilities
- →Generate documentation for features or PRs
- →Structure content into standardized sections
- →Consolidate multiple documentation files
- →Create human-readable documentation for non-technical reviewers
- →Write concise test scenarios with verification steps
How it works
The skill guides the creation of concise, human-readable documentation by enforcing a 'one document, human scale' pattern. It structures content into sections like feature summary, files changed, key concepts, installation, tests, and approval criteria.
Inputs & outputs
When to use writing-project-docs
- →Writing PR summaries for reviewers
- →Documenting feature changes for project leads
- →Consolidating fragmented documentation
- →Creating readable implementation overviews
About this skill
Writing Project Docs
Overview
Documentation must be readable by a human in a single sitting.
The default failure mode is generating exhaustive, internally consistent documentation that no one reads. The goal is a single concise document that covers everything a person needs to know, do, and verify — nothing more.
When to Use
Use this skill when:
- Generating documentation for a PR, feature, or implementation
- Existing docs are too long (multiple large files, 500+ lines, duplicated content)
- A reviewer, tester, or team lead needs to understand what was done without reading all the code
- You are consolidating multiple documentation files into one
Do NOT use for:
- Internal code comments or docstrings
- Architecture decision records (ADRs)
- API reference docs (those should be exhaustive)
Core Pattern — One Document, Human Scale
❌ BAD: 6 separate files × 300+ lines each = no one reads any of them
✅ GOOD: 1 file × ~150 lines = a person reads it in 5 minutes and knows what to do
Maximum length target: ~150–200 lines for a feature or PR doc.
If it's longer, cut it — don't add a summary on top.
Document Structure (use this order)
# [Feature name] — [One-line description]
### [System] · [Context] · [Branch/PR]
## ¿Qué hace? / What does it do?
[3–5 lines max. Problem → Solution → Benefit]
## Archivos modificados / Files changed
[Table: Layer | File | What changed — one line per file]
## [Key concept] — [any modes, states, or rules]
[Table or small diagram. No prose.]
## Cómo instalar / How to install
[Numbered steps. Commands ready to copy-paste. Verification query included.]
## Pruebas / Tests
[One subsection per scenario. Format: Who, Steps (numbered, short), Expected result, Verification SQL if needed]
## Criterios de aprobación / Approval criteria
[Two lists: Mandatory (blocking) | Recommended]
## Commit
[Copy-paste ready git command]
## Rollback
[Copy-paste ready SQL/command]
## Problemas comunes / Common issues
[Table: Problem | Cause | Fix]
Rules
Cut ruthlessly
- One doc, not six. If you produced multiple docs, merge them.
- Each section must add information not present elsewhere.
- No "executive summary" on top of a document that is already short.
- No "notes for the tester" section that repeats what's already in the test scenarios.
Tables over prose
Every time you want to write a paragraph, ask: "Is this a table?"
❌ "The system has four operating modes. In CARGA mode the user..."
✅ | Mode | When | Title | Buttons | Fields |
|------|------|-------|---------|--------|
Commands must be copy-paste ready
❌ "Run the appropriate msbuild command with Debug configuration"
✅ npm run build
Test scenarios: Who + Steps + Result + SQL
Each scenario must answer exactly four questions:
- Who is logged in?
- What steps does the tester follow? (numbered, ≤ 8 steps)
- What is the expected result? (visible behavior + message)
- How to verify in the DB? (copy-paste SQL when relevant)
Language
Write in the language of the project team. Match the team's default language for headings and explanatory text. Keep file names, commands, SQL, and code in their native syntax.
Common Mistakes
| Mistake | Fix |
|---|---|
| Generating 5+ files for one feature | Merge into one GUIA-RAPIDA PR {PR_NUMBER}.md |
| Writing "Resumen ejecutivo" + detailed sections | Choose one — the detail IS the summary |
| Listing files without saying what changed | Add a "What changed" column |
| Test scenarios without DB verification | Add the SQL query |
| Long prose in "How to install" | Convert to numbered steps with commands |
| Documenting edge cases no one will test | Move to "Known limitations" in 2 lines |
| Repeating the approval criteria in 3 places | One list, at the end |
Quick Reference
| Question | Answer |
|---|---|
| Max length? | ~150–200 lines |
| How many files? | One per feature/PR (GUIA-RAPIDA PR {PR, feature, or implementation_NUMBER}.md) |
| Language for headers? | Project team language |
| Language for code/SQL? | English |
| Test format? | Who · Steps · Result · SQL |
| Commands format? | Copy-paste ready, no placeholders |
| When to use tables? | Always, instead of prose lists |
Example — Test Scenario (good)
### Escenario 3 — Invitación de usuario ✅ *(crítico)*
**Quién:** Administrador con permisos para alta de usuarios
1. Ir a Administración → Usuarios
2. Hacer clic en **Invitar usuario**
3. Completar email y rol
4. Confirmar la acción
**Resultado esperado:**
```sql
SELECT email, status FROM user_invites
WHERE email = '[email protected]';
-- email = '[email protected]', status = 'pending'
## Real-World Impact
In one rollout, a documentation folder had 6 files totalling ~86 KB and ~2,250 lines. After applying this pattern: one file, ~200 lines, covering the test scenarios, installation steps, critical review points, commit command, and rollback. A reviewer could read it in under 5 minutes.
When not to use it
- →The task involves internal code comments or docstrings
- →The task involves architecture decision records (ADRs)
- →The task involves API reference docs
Limitations
- →Maximum length target of ~150–200 lines for a feature or PR doc
- →Not for internal code comments, docstrings, architecture decision records (ADRs), or API reference docs
- →Requires commands to be copy-paste ready
How it compares
This skill prioritizes brevity and clarity for non-technical audiences, structuring documentation into specific, concise sections, unlike exhaustive technical documentation.
Compared to similar skills
writing-project-docs side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| writing-project-docs (this skill) | 0 | 2mo | No flags | Intermediate |
| ml-paper-writing | 48 | 6mo | Review | Advanced |
| docs-review | 10 | 7mo | No flags | Beginner |
| claude-md-improver | 21 | 6mo | Review | Beginner |
Try saying
Example prompts that trigger this skill in your AI assistant.
More by Lubonch
View all by Lubonch →You might also like
ml-paper-writing
davila7
Write publication-ready ML/AI papers for NeurIPS, ICML, ICLR, ACL, AAAI, COLM. Use when drafting papers from research repos, structuring arguments, verifying citations, or preparing camera-ready submissions. Includes LaTeX templates, reviewer guidelines, and citation verification workflows.
docs-review
metabase
Review documentation changes for compliance with the Metabase writing style guide. Use when reviewing pull requests, files, or diffs containing documentation markdown files.
claude-md-improver
anthropics
Audit and improve CLAUDE.md files in repositories. Use when user asks to check, audit, update, improve, or fix CLAUDE.md files. Scans for all CLAUDE.md files, evaluates quality against templates, outputs quality report, then makes targeted updates. Also use when the user mentions "CLAUDE.md maintenance" or "project memory optimization".
write-docs
tldraw
Writing SDK documentation for tldraw. Use when creating new documentation articles, updating existing docs, or when documentation writing guidance is needed. Applies to docs in apps/docs/content/.
update-docs
vercel
This skill should be used when the user asks to "update documentation for my changes", "check docs for this PR", "what docs need updating", "sync docs with code", "scaffold docs for this feature", "document this feature", "review docs completeness", "add docs for this change", "what documentation is affected", "docs impact", or mentions "docs/", "docs/01-app", "docs/02-pages", "MDX", "documentation update", "API reference", ".mdx files". Provides guided workflow for updating Next.js documentation based on code changes.
wiki-architect
microsoft
Analyzes code repositories and generates hierarchical documentation structures with onboarding guides. Use when the user wants to create a wiki, generate documentation, map a codebase structure, or understand a project's architecture at a high level.