technical-documentation-framework
Skill2 starsChanged 4 months agoApache-2.0View on GitHub
A Claude Skill framework for generating clear, structured, example-rich, terminology-controlled technical documentation for any product, platform, service, SDK, API, workflow, or technical system — delivered as Markdown, Word (DOCX), PDF, HTML, or slides.
In this repository11 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- audience-alignmentSkill
- document-architectureSkill
- examples-and-edge-casesSkill
- glossary-builderSkill
- knowledge-mappingSkill
- output-formatsSkill
- product-discoverySkill
- publishingSkill
- section-writerSkill
- technical-reviewSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
17 rules. About 0 are common, 15 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈15
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Use this skill when the user asks for any of the following:
- Distinctive: line 23
Document this platform, service, API, SDK, library, agent, workflow, or system.
- Distinctive: line 25
Create a developer guide.
- Distinctive: line 26
Create an architecture document.
- Distinctive: line 27
Create an API reference.
- Distinctive: line 30
Create a Claude Skill for documentation generation.
- Distinctive: line 33
Update existing documentation after a product change.
- Distinctive: line 34
Write release notes, a migration guide, or a runbook.
- Distinctive: line 57
Before writing any documentation, detect the language of the user's current request and conversation.
- Very specific: line 68
Read and apply: rules/language-rules.md.
- Distinctive: line 114
Create a structured technical map of the product.
- Very specific: line 167
Read: section-writer/SKILL.md and rules/natural-writing-rules.md
- Distinctive: line 170
Write the full documentation in clean, publishable Markdown, in a voice that reads human.
- Distinctive: line 197
When not to use this product or feature
- Distinctive: line 267
Every important technical term must be explained either on first use or in the glossary.
- Distinctive: line 268
Every main section must include at least one example or scenario when applicable.
- Distinctive: line 283
When delivering the final documentation, provide:
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Technical Documentation Framework
This skill produces high-quality technical documentation for any product or technical system.
It must not behave like a simple writing prompt. It must operate as a documentation pipeline: understand the product, identify the audience, map the knowledge, define terminology, design the document structure, write the sections, add examples and edge cases, review the result, prepare the final documentation package, and deliver it in the format the user asked for.
Two standards are non-negotiable across every phase:
- The output must read as if a knowledgeable engineer wrote it. Apply
rules/natural-writing-rules.mdwhile writing, not as an afterthought. - The deliverable is what the user asked for. If they asked for a Word file with screenshots, the deliverable is a
.docxon disk with the screenshots embedded — not Markdown plus an apology.
Activation criteria
Use this skill when the user asks for any of the following:
- Write technical documentation.
- Create product documentation.
- Document this platform, service, API, SDK, library, agent, workflow, or system.
- Create a README.
- Create a developer guide.
- Create an architecture document.
- Create an API reference.
- Create a glossary.
- Turn rough product notes into documentation.
- Create a Claude Skill for documentation generation.
- Export or convert documentation to Word, PDF, HTML, or slides.
- Produce a document that includes screenshots of a product or UI.
- Update existing documentation after a product change.
- Write release notes, a migration guide, or a runbook.
Default behavior
Unless the user specifies otherwise:
| Parameter | Default |
| Documentation language | Same language as the user's current conversation |
| Tone | Clear, explanatory, technically accurate |
| Audience | Developers, technical product teams, and architects |
| Depth | Medium to advanced |
| Output format | Whatever the user requests (DOCX, PDF, HTML, slides, ...); Markdown when unspecified |
| Visual design | Brand-aware and information-first, per rules/document-design-rules.md |
| Writing voice | Natural, human, per rules/natural-writing-rules.md |
| Glossary | Required |
| Examples | Required |
| Edge cases | Required |
| Technical review | Required |
| Assumption handling | Explicitly label assumptions |
Critical language rule
Before writing any documentation, detect the language of the user's current request and conversation.
The documentation must be written in that language unless the user explicitly requests another language.
Examples:
- User speaks Turkish → write the documentation in Turkish.
- User speaks English → write the documentation in English.
- User speaks Turkish but says "write it in English" → write the documentation in English.
- User switches language and gives a clear instruction → follow the latest explicit instruction.
Read and apply: rules/language-rules.md.
Mandatory pipeline
The following steps must be followed in order.
Phase 1 — Product Discovery
Read: product-discovery/SKILL.md
Goal: Understand the product, the problem it solves, its users, components, integrations, inputs, outputs, assumptions, and boundaries.
Expected output:
- Product summary
- Problem statement
- Target users
- Main use cases
- Core components
- Integration points
- Product boundaries
- Assumptions
- Open questions
Phase 2 — Audience Alignment
Read: audience-alignment/SKILL.md
Goal: Identify who the documentation is for and how deep the explanation should be.
Expected output:
- Primary audience
- Secondary audience
- Technical level
- Required prior knowledge
- Concepts that must be explained
- Tone and terminology policy
Phase 3 — Knowledge Mapping
Read: knowledge-mapping/SKILL.md
Goal: Create a structured technical map of the product.
Expected output:
- Concept map
- Component map
- Data flows
- Runtime flows
- API or SDK surfaces
- Configuration areas
- Security concerns
- Operational concerns
- Observability points
- Failure scenarios
Phase 4 — Glossary Building
Read: glossary-builder/SKILL.md
Goal: Extract and define every important term used in the documentation.
Expected output:
- Term
- Short definition
- Detailed explanation
- Product-specific meaning
- Example
- Common confusion
- Related terms
The glossary is mandatory. Documentation is incomplete without it.
Phase 5 — Document Architecture
Read: document-architecture/SKILL.md
Goal: Design the final documentation package.
Expected output:
- Recommended document package
- Table of contents
- Section order
- Purpose of each section
- Required examples per section
- Diagram recommendations
- Cross-reference plan
Phase 6 — Section Writing
Read: section-writer/SKILL.md and rules/natural-writing-rules.md
Goal: Write the full documentation in clean, publishable Markdown, in a voice that reads human.
Expected output:
- Complete sections
- Clear headings
- Concept explanations
- Usage flows
- Configuration guidance
- Integration guidance
- API or SDK explanations when relevant
Phase 7 — Examples and Edge Cases
Read: examples-and-edge-cases/SKILL.md
Goal: Make the documentation practical and realistic.
Expected output:
- Simple example
- Realistic example
- Incorrect usage example
- Edge-case example
- Decision notes
- Common mistakes
- When not to use this product or feature
Phase 8 — Technical Review
Read: technical-review/SKILL.md
Goal: Review documentation for correctness, clarity, depth, example quality, and terminology consistency.
Expected output:
- QA score
- Missing sections
- Inconsistent terms
- Shallow areas
- Risky assumptions
- Recommended revisions
- Final readiness decision
Phase 9 — Publishing
Read: publishing/SKILL.md
Goal: Prepare the final documentation package.
Expected output:
docs/
├── README.md
├── overview.md
├── getting-started.md
├── concepts.md
├── architecture.md
├── configuration.md
├── api-reference.md
├── examples.md
├── troubleshooting.md
├── glossary.md
└── faq.mdPhase 10 — Format and Delivery
Read: output-formats/SKILL.md and rules/document-design-rules.md
Goal: Deliver the reviewed documentation in the format the user requested, with a visual design that is professional, information-first, and branded to the company the document belongs to.
Expected output:
- The final artifact in the requested format (DOCX, PDF, HTML, slides, or Markdown), written to disk and verified
- For DOCX/PDF: brand palette researched from the owning company's official identity materials; cover page with metadata table; TOC field; part dividers and numbered chapters; callout boxes, branded tables, styled code blocks; figures renumbered in reading order with a figure index appendix
- Screenshots or rendered diagrams embedded where the content calls for them
- Schema validation passed and rendered sample pages visually inspected; when redesigning an existing document, a content-loss probe run against the source
- The source Markdown package alongside the converted artifact, unless the user declines it
Skip the conversion step only when Markdown itself is the requested deliverable. Never skip the verification step: an artifact that was not opened and checked is not delivered.
Updating existing documentation
When the request is to update documentation rather than create it, do not rerun the full pipeline blindly. Read and apply rules/maintenance-rules.md: work diff-driven from what actually changed in the product, touch every place the old behavior is stated, match the surrounding voice, and re-run technical review scoped to the changed documents. Release notes, migration guides, and deprecation notes follow the same rules and have templates in templates/.
Reference output
examples/sample-output-signalhop/ is a complete documentation package for a fictional product, written to this framework's standards. Use it as the quality bar during section writing and technical review — especially for voice, example density, boundary statements, and glossary depth.
Mandatory quality rules
- Every important technical term must be explained either on first use or in the glossary.
- Every main section must include at least one example or scenario when applicable.
- Product boundaries must be explicit.
- A "What this product is not" or equivalent section must exist.
- The reader must understand not only what the product does, but why it exists and how it behaves.
- The tone must be technical and explanatory, not promotional.
- Unknown information must not be invented. It must be labeled as an assumption or open question.
- The glossary must guide terminology across the whole documentation.
- Terminology inconsistency is a QA failure.
- The final output must be publishable in the requested format; the working format is clean Markdown.
- Documentation must be written in the language of the user's conversation unless explicitly overridden.
- The text must pass the natural-writing self-check in
rules/natural-writing-rules.md. Documentation that reads machine-generated is a QA failure regardless of technical accuracy. - Formatted deliverables (DOCX, PDF) must pass the design self-check in
rules/document-design-rules.md: brand-sourced palette, complete document anatomy, information rendered visible (tables, callouts, figure index), and a verified, content-complete artifact. Default styling on a formatted deliverable is a QA failure.
Final response behavior
When delivering the final documentation, provide:
- A short summary of what was produced.
- The documentation package structure.
- The final artifact in the requested format, plus the source Markdown package.
- Any assumptions made.
- QA result, including the natural-writing check.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

