
document-generation-pdf
Generate, fill, and assemble PDF documents at scale. Handles legal forms, contracts, invoices, certificates. Supports form filling (pdf-lib), template rendering (Puppeteer, LaTeX), digital signatures (DocuSign), and document assembly. Use for legal tech, HR automation, invoice generation. Activate on "PDF generation", "form filling", "document automation", "digital signatures". NOT for simple PDF viewing, basic file conversion, or OCR text extraction.
- Installs
- 273
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/curiositech/some_claude_skills --skill document-generation-pdf- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads document-generation-pdf automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
Document Generation & PDF Automation
Expert in generating, filling, and assembling PDF documents programmatically for legal, HR, and business workflows.
When to Use
✅ Use for:
- Legal form automation (expungement, immigration, contracts)
- Invoice/receipt generation at scale
- Certificate creation (completion, participation, awards)
- Contract assembly from templates
- Government form filling (IRS, court filings)
- Multi-document packet creation
❌ NOT for:
- Simple PDF viewing (use browser or pdf.js)
- Basic file conversion (use online tools)
- OCR text extraction (use Tesseract.js or AWS Textract)
- PDF editing by hand (use Adobe Acrobat)
Technology Selection
pdf-lib vs Puppeteer vs LaTeX
| Feature | pdf-lib | Puppeteer | LaTeX |
|---|---|---|---|
| Form filling | ✅ Native | ❌ Complex | ❌ No |
| Template rendering | ❌ No | ✅ HTML/CSS | ✅ Templates |
| Performance (1000 PDFs) | 5s | 60s | 30s |
| File size | Small | Medium | Small |
| Signature fields | ✅ Yes | ❌ No | ❌ No |
| Best for | Government forms | Invoices, reports | Academic papers |
Timeline:
- 2000s: LaTeX for academic documents
- 2010: PDFKit (Node.js) for generation
- 2017: Puppeteer for HTML → PDF
- 2019: pdf-lib for pure JS form filling
- 2024: pdf-lib is gold standard for forms
Decision tree:
Need to fill existing form? → pdf-lib
Need complex layouts? → Puppeteer (HTML/CSS)
Need academic formatting? → LaTeX
Need to merge PDFs? → pdf-lib
Need digital signatures? → pdf-lib + DocuSign API
Common Anti-Patterns
Anti-Pattern 1: Using Puppeteer for Simple Form Filling
Novice thinking: "I'll use Puppeteer for everything, it's versatile"
Problem: 12x slower, 10x more memory, can't preserve form fields.
Wrong approach:
// ❌ Puppeteer for simple form filling (SLOW!)
import puppeteer from 'puppeteer';
async function fillForm(data: FormData): Promise<Buffer> {
const browser = await puppeteer.launch();
const page = await browser.newPage();
// Load PDF in browser
await page.goto(`file://${pdfPath}`);
// Somehow fill form fields? (hacky)
await page.evaluate((data) => {
// Can't easily access PDF form fields from DOM
// Would need to convert PDF → HTML first
}, data);
const pdf = await page.pdf();
await browser.close();
return pdf;
}
Why wrong:
- Browser overhead (200MB+ RAM per instance)
- Can't access native PDF form fields
- Loses interactive form capabilities
- 12x slower than pdf-lib
Correct approach:
// ✅ pdf-lib for form filling (FAST!)
import { PDFDocument } from 'pdf-lib';
async function fillForm(templatePath: string, data: FormData): Promise<Uint8Array> {
// Load existing PDF form
const existingPdfBytes = await fs.readFile(templatePath);
const pdfDoc = await PDFDocument.load(existingPdfBytes);
// Get form
const form = pdfDoc.getForm();
// Fill text fields
form.getTextField('applicant_name').setText(data.name);
form.getTextField('case_number').setText(data.caseNumber);
form.getTextField('date_of_birth').setText(data.dob);
// Fill checkboxes
if (data.hasPriorConvictions) {
form.getCheckBox('prior_convictions').check();
}
// Fill dropdowns
form.getDropdown('state').select(data.state);
// Flatten form (make non-editable)
form.flatten();
// Save
return await pdfDoc.save();
}
Performance comparison (1000 PDFs):
- Puppeteer: 60 seconds, 4GB RAM
- pdf-lib: 5 seconds, 200MB RAM
Timeline context:
- 2017: Puppeteer released, everyone used it for PDFs
- 2019: pdf-lib released, proper form handling
- 2024: pdf-lib is standard for government forms
Anti-Pattern 2: Not Flattening Forms
Problem: Users can edit filled forms, causing data inconsistencies.
Wrong approach:
// ❌ Don't flatten - form stays editable
const pdfDoc = await PDFDocument.load(existingPdfBytes);
const form = pdfDoc.getForm();
form.getTextField('name').setText('John Doe');
const pdfBytes = await pdfDoc.save();
// User can open PDF and change "John Doe" to anything!
Why wrong:
- User can modify official documents
- Data doesn't match database
- Violates document integrity
Correct approach:
// ✅ Flatten form after filling
const pdfDoc = await PDFDocument.load(existingPdfBytes);
const form = pdfDoc.getForm();
form.getTextField('name').setText('John Doe');
// Flatten (convert fields to static text)
form.flatten();
const pdfBytes = await pdfDoc.save();
// User can't edit filled values ✅
When NOT to flatten:
- Draft documents (user needs to review/edit)
- Multi-step workflows (partial completion)
- Templates for users to fill manually
Anti-Pattern 3: Generating PDFs from HTML Without Page Breaks
Novice thinking: "HTML → PDF is easy with Puppeteer"
Problem: Content splits mid-sentence across pages.
Wrong approach:
// ❌ No page break control
const html = `
<div class="contract">
<h1>Mutual Agreement</h1>
<p>Long paragraph that might split across pages...</p>
<section>
<h2>Terms and Conditions</h2>
<ol>
<li>Term 1 that could get cut off...</li>
<li>Term 2...</li>
</ol>
</section>
</div>
`;
const pdf = await page.pdf({ format: 'A4' });
// Result: Ugly page breaks in middle of sections
Correct approach:
// ✅ Explicit page break control with CSS
const html = `
<style>
@media print {
.page-break { page-break-after: always; }
.avoid-break { page-break-inside: avoid; }
h1, h2, h3 {
page-break-after: avoid;
page-break-inside: avoid;
}
section {
page-break-inside: avoid;
}
}
</style>
<div class="contract">
<section class="avoid-break">
<h1>Mutual Agreement</h1>
<p>This entire section stays together...</p>
</section>
<div class="page-break"></div>
<section class="avoid-break">
<h2>Terms and Conditions</h2>
<ol>
<li>Term 1</li>
<li>Term 2</li>
</ol>
</section>
</div>
`;
const pdf = await page.pdf({
format: 'A4',
printBackground: true,
margin: {
top: '1in',
right: '1in',
bottom: '1in',
left: '1in'
}
});
CSS print properties:
page-break-before: always- Force new page before elementpage-break-after: always- Force new page after elementpage-break-inside: avoid- Keep element together
Anti-Pattern 4: Not Handling Signature Fields
Problem: Signature fields aren't clickable in generated PDFs.
Wrong approach:
// ❌ Add signature as image (not a real signature field)
const pdfDoc = await PDFDocument.load(existingPdfBytes);
const signatureImage = await pdfDoc.embedPng(signaturePngBytes);
const page = pdfDoc.getPage(0);
page.drawImage(signatureImage, {
x: 100,
y: 100,
width: 200,
height: 50
});
await pdfDoc.save();
// Not a real signature field - can't be signed digitally
Why wrong:
- Not legally recognized (just an image)
- Can't use DocuSign/Adobe Sign
- No signature metadata (who, when)
Correct approach 1: Create signature field (for DocuSign)
// ✅ Create signature field for electronic signing
const pdfDoc = await PDFDocument.load(existingPdfBytes);
const form = pdfDoc.getForm();
// Create signature field
const signatureField = form.createTextField('applicant_signature');
signatureField.addToPage(pdfDoc.getPage(0), {
x: 100,
y: 100,
width: 200,
height: 50
});
// Mark as signature field (metadata)
signatureField.updateWidgets({
borderWidth: 1,
borderColor: { type: 'RGB', red: 0, green: 0, blue: 0 }
});
await pdfDoc.save();
// DocuSign can now detect and fill this field ✅
Correct approach 2: DocuSign API integration
// ✅ Send to DocuSign for e-signature
import { ApiClient, EnvelopesApi } from 'docusign-esign';
async function sendForSignature(pdfBytes: Uint8Array, signerEmail: string) {
const apiClient = new ApiClient();
apiClient.setBasePath('https://demo.docusign.net/restapi');
const envelopesApi = new EnvelopesApi(apiClient);
const envelope = {
emailSubject: 'Please sign: Expungement Petition',
documents: [{
documentBase64: Buffer.from(pdfBytes).toString('base64'),
name: 'Petition.pdf',
fileExtension: 'pdf',
documentId: '1'
}],
recipients: {
signers: [{
email: signerEmail,
name: 'John Doe',
recipientId: '1',
tabs: {
signHereTabs: [{
documentId: '1',
pageNumber: '1',
xPosition: '100',
yPosition: '100'
}]
}
}]
},
status: 'sent'
};
return await envelopesApi.createEnvelope(accountId, { envelopeDefinition: envelope });
}
Anti-Pattern 5: Storing Filled PDFs Without Encryption
Problem: Sensitive legal documents stored in plain text.
Wrong approach:
// ❌ Save PDF to disk unencrypted
const pdfBytes = await pdfDoc.save();
await fs.writeFile(`./documents/${caseId}.pdf`, pdfBytes);
// Sensitive data accessible to anyone with file system access
Why wrong:
- Legal/medical data exposed
- HIPAA/GDPR violations
- Data breach liability
Correct approach 1: Encrypt at rest
// ✅ Encrypt PDF with user password
const pdfBytes = await pdfDoc.save({
userPassword: generateSecurePassword(),
ownerPassword: process.env.PDF_OWNER_PASSWORD,
permissions: {
printing: 'highResolution',
modifying: false,
copying: false,
annotating: false,
fillingForms: false,
contentAccessibility: true,
documentAssembly: false
}
});
await fs.writeFile(`./documents/${caseId}.pdf`, pdfBytes);
Correct approach 2: Store in encrypted storage (S3 with SSE)
// ✅ Upload to S3 with server-side encryption
import { S3Client, PutObjectCommand } from '@aws-sdk/client-s3';
const s3Client = new S3Client({ region: 'us-east-1' });
await s3Client.send(new PutObjectCommand({
Bucket: 'expungement-documents',
Key: `cases/${caseId}/petition.pdf`,
Body: pdfBytes,
ServerSideEncryption: 'AES256',
Metadata: {
caseId: caseId,
documentType: 'petition',
generatedAt: new Date().toISOString()
},
ACL: 'private' // Not publicly accessible
}));
// Store S3 URL in database (encrypted)
await db.documents.insert({
case_id: caseId,
s3_url: encrypt(`s3://expungement-documents/cases/${caseId}/petition.pdf`),
created_at: new Date()
});
Compliance requirements:
- HIPAA: Encryption at rest + in transit, access logging
- GDPR: Right to deletion, data minimization
- State bar rules: Attorney-client privilege protection
Production Checklist
□ Form fields filled correctly (test with validation)
□ Forms flattened after filling (non-editable)
□ Page breaks controlled (no mid-sentence splits)
□ Signature fields created (for DocuSign/Adobe Sign)
□ PDFs encrypted at rest (S3 SSE or user password)
□ Access logged (who viewed/downloaded)
□ Auto-deletion scheduled (retention policy)
□ Fonts embedded (cross-platform compatibility)
□ File size optimized (<5MB per document)
□ Batch generation tested (1000+ PDFs)
When to Use vs Avoid
| Scenario | Appropriate? |
|---|---|
| Fill 50+ government forms | ✅ Yes - automate with pdf-lib |
| Generate invoices from template | ✅ Yes - Puppeteer from HTML |
| Create certificates at scale | ✅ Yes - pdf-lib or LaTeX |
| Assemble multi-doc packets | ✅ Yes - pdf-lib merge |
| View PDFs in browser | ❌ No - use pdf.js or browser |
| Edit PDF by hand | ❌ No - use Adobe Acrobat |
| Extract text with OCR | ❌ No - use Tesseract/Textract |
Flat PDF Overlay at Scale (Government Forms)
Many government court forms are flat PDFs (no fillable fields). You MUST overlay text at precise X,Y coordinates. Never generate forms from scratch - courts will reject them.
The Three-Tier Strategy
1. FILLABLE PDF → Use form.getTextField().setText() (best)
2. FLAT PDF + COORDINATES → Draw text overlays at X,Y positions (this section)
3. NO TEMPLATE EXISTS → Emergency fallback with warning banner (worst)
Anti-Pattern 6: Generating Forms from Scratch
Novice thinking: "The PDF doesn't have form fields, I'll just create a new PDF"
Problem: Courts reject non-official forms. Your custom layout won't match official documents.
Wrong approach:
// ❌ NEVER DO THIS - courts reject custom forms
const pdfDoc = await PDFDocument.create();
const page = pdfDoc.addPage([612, 792]);
page.drawText('IN THE CIRCUIT COURT OF...', { x: 200, y: 720 });
page.drawText(`Defendant: ${data.name}`, { x: 100, y: 600 });
// Result: Court clerk says "This isn't our form" and rejects filing
Correct approach: Draw on top of official template
// ✅ Load official form, draw text in blank spaces
const templateBytes = await fs.readFile('public/forms/OR-set-aside-motion.pdf');
const pdfDoc = await PDFDocument.load(templateBytes);
const page = pdfDoc.getPage(3); // Page 4 is the form
const font = await pdfDoc.embedFont(StandardFonts.Helvetica);
// Draw in blank spaces at measured coordinates
page.drawText(data.county.toUpperCase(), { x: 310, y: 700, size: 11, font });
page.drawText(data.caseNumber, { x: 432, y: 655, size: 10, font });
page.drawText(data.fullName, { x: 72, y: 568, size: 11, font });
Measuring Coordinates at Scale
Problem: Manually measuring X,Y for 50 states × 10+ forms = impossible.
Solution 1: Visual Coordinate Picker Tool
Use pdf-coordinates - click on PDF to capture X,Y:
- Upload PDF, click on blank fields
- Select coordinate system: Bottom-Left (PDF standard)
- Export annotated PDF with coordinate labels
- Works offline, no data upload
Solution 2: JSON Configuration (Zerodha Pattern)
Define coordinates in JSON config, not code:
// field-mappings/oregon-set-aside.json
{
"templatePath": "/forms/or/OR-criminal-set-aside.pdf",
"pageIndex": 3,
"fields": [
{
"name": "county",
"x": 310,
"y": 700,
"fontSize": 11,
"maxWidth": 120,
"transform": "uppercase"
},
{
"name": "caseNumber",
"x": 432,
"y": 655,
"fontSize": 10,
"maxWidth": 100,
"maxChars": 12
},
{
"name": "fullName",
"x": 72,
"y": 568,
"fontSize": 11,
"maxWidth": 250,
"shrinkToFit": true
}
]
}
Solution 3: Debug Grid Overlay
Add temporary grid during development:
async function drawDebugGrid(page: PDFPage, font: PDFFont) {
const { width, height } = page.getSize();
// Draw grid every 50 points
for (let x = 0; x <= width; x += 50) {
page.drawLine({
start: { x, y: 0 },
end: { x, y: height },
thickness: 0.2,
color: rgb(0.9, 0.9, 0.9),
});
if (x % 100 === 0) {
page.drawText(String(x), { x: x + 2, y: 5, size: 6, font });
}
}
// Same for horizontal lines...
}
Text Overflow Handling
Problem: Long names overflow into adjacent fields or get cut off.
Three overflow strategies:
// 1. Truncate with ellipsis
function truncate(text: string, maxChars: number): string {
if (text.length <= maxChars) return text;
return text.slice(0, maxChars - 2) + '..';
}
// 2. Shrink font to fit (min 7pt for readability)
function shrinkToFit(text: string, maxWidthPts: number, startSize: number): number {
let fontSize = startSize;
const charWidth = (size: number) => size * 0.52; // Helvetica avg
while (fontSize > 7 && text.length * charWidth(fontSize) > maxWidthPts) {
fontSize -= 0.5;
}
return fontSize;
}
// 3. Multi-line wrap (for addresses)
page.drawText(longAddress, {
x: 100,
y: 500,
size: 9,
font,
maxWidth: 200, // pdf-lib auto-wraps at this width
lineHeight: 12,
});
Field input constraints - Prevent overflow at data collection:
export const FIELD_CONSTRAINTS = {
fullName: { maxLength: 35 },
county: { maxLength: 12 },
caseNumber: { maxLength: 12 },
agency: { maxLength: 28 },
charges: { maxLength: 45 },
} as const;
Production Workflow for Multi-State Forms
1. DOWNLOAD official forms
└─ Store in /public/forms/{state}/ with metadata.json
2. INSPECT each form
└─ Run pdf-lib inspection to check: fillable? how many fields?
3. CATEGORIZE forms
├─ Fillable → Map field names in field-mappings/{state}.ts
└─ Flat → Measure coordinates with pdf-coordinates tool
4. CREATE JSON config for flat PDFs
└─ One JSON file per form with all X,Y coordinates
5. TEST with debug grid
└─ Generate test PDF with colored text + coordinate labels
6. QA visual verification
└─ Compare filled PDF against original blank form
7. DOCUMENT constraints
└─ Export maxLength/maxWidth limits for form inputs
Legal Filing Requirements
Court e-filing systems require:
- ✅ Flattened forms (no interactive fields)
- ✅ No digital signature metadata (DocuSign breaks e-filing)
- ✅ Standard fonts embedded (Helvetica, Times)
- ✅ File under 10MB
- ✅ Exact match to official form layout
Flattening for courts:
// If form had fields, flatten them
const form = pdfDoc.getForm();
if (form.getFields().length > 0) {
form.flatten(); // Converts to static text
}
// For overlay text, it's already static - no flattening needed
const pdfBytes = await pdfDoc.save();
Coordinate System Reference
PDF coordinate system (pdf-lib default):
- Origin: Bottom-left corner (0, 0)
- X increases: Left → Right
- Y increases: Bottom → Top
- Units: Points (72 points = 1 inch)
- Letter size: 612 x 792 points (8.5" x 11")
To convert from top-left origin (e.g., Adobe Acrobat display):
y_pdf = 792 - y_topLeft
Common positions (Letter size):
- Top margin: y = 720-750
- Header area: y = 700-750
- Body start: y = 650-700
- Left margin: x = 50-72
- Right edge: x = 540-560
- Bottom margin: y = 50-72
References
/references/pdf-lib-guide.md- Form filling, field types, flattening, encryption/references/puppeteer-templates.md- HTML templates, page breaks, styling for print/references/document-assembly.md- Merging PDFs, packet creation, watermarks- pdf-coordinates - Visual X,Y coordinate picker
- Zerodha pdf_text_overlay - JSON config pattern
Scripts
scripts/form_filler.ts- Fill PDF forms from JSON data, batch processingscripts/document_assembler.ts- Merge multiple PDFs, add cover pages, watermarksscripts/generate-test-overlay-pdf.ts- Test overlay coordinates with debug grid
This skill guides: PDF generation | Form filling | Document automation | Digital signatures | pdf-lib | Puppeteer | LaTeX | DocuSign | Flat PDF overlay | Coordinate mapping
FAQ
What does the document-generation-pdf skill do?
Generate, fill, and assemble PDF documents at scale. Handles legal forms, contracts, invoices, certificates. Supports form filling (pdf-lib), template rendering (Puppeteer, LaTeX), digital signatures (DocuSign), and document assembly. Use for legal tech, HR automation, invoice generation. Activate on "PDF generation", "form filling", "document automation", "digital signatures". NOT for simple PDF viewing, basic file conversion, or OCR text extraction.
How do I install document-generation-pdf?
Run npx skills add https://github.com/curiositech/some_claude_skills --skill document-generation-pdf in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/document-generation-pdf/ (personal) or .claude/skills/document-generation-pdf/ (project).
Which AI agents can use document-generation-pdf?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is document-generation-pdf safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains document-generation-pdf?
It is published by curiositech in curiositech/some_claude_skills. It has 273 recorded installs.
More from curiositech/some_claude_skills
video-processing-editing
curiositech/some_claude_skills
FFmpeg automation for cutting, trimming, concatenating videos. Audio mixing, timeline editing, transitions, effects. Export optimization for YouTube, social media. Subtitle handling, color grading, batch processing. Use for videogen projects, content creation, automated video production. Activate on "video editing", "FFmpeg", "trim video", "concatenate", "transitions", "export optimization". NOT for real-time video editing UI, 3D compositing, or motion graphics.
interior-design-expert
curiositech/some_claude_skills
Expert interior designer with deep knowledge of space planning, color theory (Munsell, NCS), lighting design (IES standards), furniture proportions, and AI-assisted visualization. Use for room layout optimization, lighting calculations, color palette selection for interiors, furniture placement, style consultation. Activate on "interior design", "room layout", "lighting design", "furniture placement", "space planning", "Munsell color". NOT for exterior/landscape design, architectural structure, web/UI design (use web-design-expert), brand color theory (use color-theory-palette-harmony-expert), or building codes/permits.
cv-creator
curiositech/some_claude_skills
Professional CV and resume builder transforming career narratives into ATS-optimized, multi-format resumes. Integrates with career-biographer for data and competitive-cartographer for positioning. Generates PDF, DOCX, LaTeX, JSON Resume, HTML, and Markdown. Activate on 'resume', 'CV', 'ATS optimization', 'job application'. NOT for cover letters, portfolio websites (use web-design-expert), LinkedIn optimization, or interview preparation.
personal-finance-coach
curiositech/some_claude_skills
Expert personal finance coach with deep knowledge of tax optimization, investment theory (MPT, factor investing), retirement mathematics (Trinity Study, SWR research), and wealth-building strategies grounded in academic research. Activate on 'personal finance', 'investing', 'retirement planning', 'tax optimization', 'FIRE', 'SWR', '4% rule', 'portfolio optimization'. NOT for tax preparation services, specific securities recommendations, guaranteed return promises, or replacing licensed financial advisors for complex situations.
photo-composition-critic
curiositech/some_claude_skills
Expert photography composition critic grounded in graduate-level visual aesthetics education, computational aesthetics research (AVA, NIMA, LAION-Aesthetics, VisualQuality-R1), and professional image analysis with custom tooling. Use for image quality assessment, composition analysis, aesthetic scoring, photo critique. Activate on "photo critique", "composition analysis", "image aesthetics", "NIMA", "AVA dataset", "visual quality". NOT for photo editing/retouching (use native-app-designer), generating images (use Stability AI directly), or basic image processing (use clip-aware-embeddings).
pwa-expert
curiositech/some_claude_skills
Progressive Web App development with Service Workers, offline support, and app-like behavior. Use for caching strategies, install prompts, push notifications, background sync. Activate on "PWA", "Service Worker", "offline", "install prompt", "beforeinstallprompt", "manifest.json", "workbox", "cache-first". NOT for native app development (use React Native), general web performance (use performance docs), or server-side rendering.
Popular docs & writing skills
grill-with-docs
mattpocock/skills
A relentless interview to sharpen a plan or design, which also creates docs (ADR's and glossary) as we go.
lark-markdown
open.feishu.cn
飞书 Markdown:查看、创建、上传、编辑和比较飞书中的原生 Markdown 文件。当用户要操作飞书 Markdown 文件,或比较其远端版本及本地草稿时使用。纯本地 Markdown 文件操作不触发本 skill。不负责将 Markdown 导入为飞书在线文档,也不负责文件搜索、权限、评论、移动、删除等云空间管理操作。
writing-beats
mattpocock/skills
Writing, exploit; assemble raw material into a journey of beats, grounding each term before a beat leans on it.
writing-shape
mattpocock/skills
Writing, exploit: shape raw material into an article, paragraph by paragraph.
writing-fragments
mattpocock/skills
Writing, explore: mine raw fragments, no structure yet.
writing-for-agents
mattpocock/skills
Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.