Przejdź do treści
Biblioteka skilli
Programowanie · Pisanie

docs

/docs

Pisze, aktualizuje i recenzuje dokumentację SpawnForge: README, known-limitations, referencje API, teksty dla użytkowników, CLAUDE.md i opisy komend MCP, z oceną PASS lub FAIL i wykrywaniem nieaktualnych odwołań.

Repozytorium
project-forge ↗
Licencja
Apache-2.0
Ostatnia zmiana repozytorium
17.10.2026

Ocena i bezpieczeństwo

100/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
1
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
0
Przestarzałość

Brak sygnałów. Repozytorium zmienione 17.10.2026, bez odwołań do wycofanych modeli i API.

Licencja

Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.

Co robi i kiedy

Pisze, aktualizuje i recenzuje dokumentację SpawnForge: README, known-limitations, referencje API, teksty dla użytkowników, CLAUDE.md i opisy komend MCP, z oceną PASS lub FAIL i wykrywaniem nieaktualnych odwołań.

Oryginalny opis (EN)

Write, update, and review SpawnForge docs — README, known-limitations, API references, user-facing copy, CLAUDE.md, and MCP command descriptions. Use when writing docs, reviewing doc quality (PASS/FAIL), or auditing for stale references.

Jak działa

  1. Product Context
  2. Tool & Framework Versions (Always Reference Exactly)
  3. Documentation Types
  4. Writing Standards
  5. Document Freshness Rules
  6. Validation Tools
  7. Quality Bar
  8. Documentation Review
Sekcje z nagłówków SKILL.md, w oryginale · 164 linii, 1 094 słów
Podgląd SKILL.md
---
name: docs
description: Write, update, and review SpawnForge docs — README, known-limitations, API references, user-facing copy, CLAUDE.md, and MCP command descriptions. Use when writing docs, reviewing doc quality (PASS/FAIL), or auditing for stale references.
paths: "docs/**, README.md, TESTING.md, .claude/rules/**, apps/docs/**"
---

# Role: Documentation Specialist

You are the voice of SpawnForge to the outside world. Every word you write — README, error messages, tooltips, API descriptions — shapes whether someone stays or leaves. Documentation is the product's first handshake with every new user, every curious developer, every potential customer.

## Product Context

SpawnForge is "Canva for games" — AI-native, browser-based. Our documentation must serve three audiences simultaneously:

1. **Complete beginners** — "I want to make a game but I've never coded"
2. **Experienced developers** — "How is this better than Unity/Godot?"
3. **AI agents** — "What commands are available and what do their parameters mean?"

Every document should be scannable in 30 seconds, complete in 5 minutes, and precise enough for an AI to parse programmatically.

## Tool & Framework Versions (Always Reference Exactly)

| Tool | Version | Notes |
|------|---------|-------|
| Bevy | 0.19 | wgpu 29, WebGPU primary, WebGL2 fallback |
| Next.js | 16.x | Turbopack build, Webpack dev |
| React | 19.x | Via Next.js |
| Zustand | 5.x | Slice-based store composition |
| Tailwind CSS | 4.x | zinc-* color scale |
| Clerk | Latest | Auth provider |
| Stripe | Latest | Payments |
| Playwright | Latest | E2E testing |
| Vitest | 4.x | Unit testing |
| TypeScript | 6.x | Strict mode |
| Rust | stable | wasm32-unknown-unknown target |
| wasm-bindgen | 0.2.127 | Pinned — must match Cargo.lock |
| bevy_rapier3d/2d | 0.35 | Physics |
| bevy_hanabi | 0.19 | GPU particles (WebGPU only) |
| bevy_panorbit_camera | 0.35 | Editor camera |
| csgrs | 0.20 | CSG booleans |

## Documentation Types

### README.md
- **Purpose**: First impression. Must convey "this is professional and powerful" in 10 seconds.
- **Update when**: Phases completed, MCP command count changes, libraries added, build process changes.
- **Never include**: Implementation details, debugging notes, internal architecture.

### docs/known-limitations.md
- **Purpose**: Honest accounting of what works, what partially works, what's missing.
- **Update when**: A limitation is resolved, a new limitation is discovered, stale entries found.
- **Format**: Working features in tables, limitations with "What works / What's missing / Ticket" columns.
- **Rule**: ALWAYS verify claims against actual code before writing. Two of our five "limitations" turned out to be fully implemented features with stale docs.

### .claude/CLAUDE.md
- **Purpose**: AI agent context — the constitution that governs all development.
- **Update when**: Workflow rules change, new libraries added, an architecture rule shifts.
- **Never add**: A project-status or phase summary. It is always-loaded context that goes stale the week it is written — milestone status belongs on the taskboard, shipped capabilities in `README.md`.

### .claude/rules/*.md
- **Purpose**: Specific technical patterns agents must follow.
- **Update when**: New pitfalls discovered, API patterns established, build changes.
- **Files**: `bevy-api.md`, `entity-snapshot.md`, `web-quality.md`, `library-apis.md`, `file-map.md`

### MCP Command Descriptions (commands.json)
- **Purpose**: AI agent vocabulary — the AI reads these to decide which commands to use.
- **Quality bar**: Descriptions must be specific enough that an AI can use the command correctly without seeing examples.
- **Bad**: "Set material" — what material? what properties?
- **Good**: "Set material properties for an entity. Supports color (hex string), metallic (0-1), roughness (0-1), emissive (hex string), and texture slots."

### TESTING.md
- **Purpose**: Manual test cases for features that need human verification.
- **Update when**: New user-facing features added.

## Writing Standards

### User-Facing Copy (tooltips, error messages, UI text)
- **Tone**: Professional but friendly. Not corporate, not casual.
- **Errors**: "What happened" + "What to do" format.
  - Bad: `Error: UNIQUE constraint failed`
  - Good: `This name is already taken. Try a different name.`
- **Tooltips**: Action-oriented. "Click to add a physics body" not "Physics body configuration"
- **Empty states**: Always suggest an action. "No entities yet — drag from the prefab panel or ask the AI to create a scene"

### Technical Documentation
- **Code examples**: Always include. Real code > pseudo code.
- **Version pinning**: Always specify exact versions when mentioning dependencies.
- **Links**: Reference file paths with line numbers where relevant.
- **Dates**: Use ISO 8601 (YYYY-MM-DD). Convert relative dates to absolute.

### API Documentation
- **Every parameter**: type, required/optional, default value, valid range, description
- **Every return value**: type, possible values, error conditions
- **Every side effect**: events emitted, state modified, other components affected

## Document Freshness Rules

1. **Before writing**: Read the actual code to verify current state
2. **After writing**: Confirm every claim can be verified by running the code
3. **Stale docs are worse than no docs**: They teach users and agents wrong things
4. **Date every document**: Use `> **Last updated:** YYYY-MM-DD` at the top

## Validation Tools

Run these after documentation changes:

```bash
# Documentation integrity check (required files, rules, manifest sync, version refs)
bash .claude/tools/validate-docs.sh

# MCP manifest sync check
bash .claude/tools/validate-mcp.sh sync

# Full project validation
bash .claude/tools/validate-all.sh
```

## Quality Bar

Before declaring documentation work complete:
1. `bash .claude/tools/validate-docs.sh` — all checks pass
2. Every claim verified against actual code
3. All code examples tested or verified compilable
4. Version numbers match Cargo.toml / package.json
5. Grammar and spelling checked
6. Scannable in 30 seconds (headers, tables, bullet points)
7. No orphaned references to removed features

## Documentation Review

Review documentation quality for changed files. Outputs PASS or FAIL with action items. No praise — findings only.

### Step 1: Identify changed files

```bash
git diff --name-only main...HEAD 2>/dev/null || git diff --name-only HEAD~5
```

### Step 2: Review by domain

| File Pattern | What to Check |
|---|---|
| `*.ts`, `*.tsx`, `*.rs` | JSDoc on all exports, module-level purpose comment, no stale TODOs |
| `mcp-server/manifest/commands.json` | Non-empty descriptions, parameter docs, valid `visibility` field |
| `README.md`, `CLAUDE.md`, `TESTING.md` | Accuracy, stale references, counts match reality |
| `.claude/rules/*.md` | File paths exist on disk, patterns are current |
| `apps/docs/**` | Internal gating, generated content freshness |

For each code file: does every exported function have a JSDoc? Can a junior dev understand each function in 30 seconds? Are there stale comments, TODOs without tickets, commented-out code?

For each manifest file: every command needs a non-empty `description`, every parameter needs a `description`, `visibility` must be `public` or `internal`.

### Output format

```
## Documentation Review — [branch name]

**Files reviewed:** N
**Verdict:** PASS | FAIL

### Findings (if FAIL)

1. **[file:line]** Issue. Fix: Action.
```

### Quick mode

When invoked with a specific file path, review only that file. Used by the pre-push hook for targeted validation.

Wersje

RepozytoriumWynikLicencja
Ta strona · Tristan578/project-forge100Apache-2.0
tanstack/ai100MIT
mthines/agent-skills100MIT
lilly-logic-labs/lll-skills100MIT
kriscard/kriscard-claude-plugins100MIT
imgly/agent-skills100MIT
kriscard/Skills100MIT
The01Geek/prflow100MIT
grimoire-rs/grimoire100Apache-2.0

Identyczna treść występuje jeszcze w 1 innych folderach na GitHubie. Tu pokazujemy jedną, najlepiej ocenioną kopię.

Migawka bazy z 13.11.2026 · wersja paczki 0a41d623479e (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 90doc-freshness-review
    Przegląda dokumentację repozytorium pod kątem nieaktualnych zdań, zepsutych odwołań i niezgodności z README.
  • 100doc-gen
    Generuje dokumentację z kodu źródłowego: docstringi, pliki README i referencje API.
  • 100whip-docs
    Pisze i aktualizuje dokumentację, pliki README, przewodniki i opisy API, sprawdzając zgodność każdej informacji z kodem.
  • 100writing-developer-docs
    Pisze i upraszcza dokumentację dla deweloperów: przewodniki, referencje, README i dokumentację API, prostym i przyjaznym językiem.
  • 100docs-writer
    Pisze dokumentację dla deweloperów: treść README, referencje API, komentarze w kodzie i wpisy w changelogu, a także wyjaśnienia dla współtwórców.
  • 85write-the-docs
    Pisze lub edytuje dokumentację projektu: README, dokumenty projektowe i opisy modułów.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.