Documentation
Use these patterns when writing documentation, creating runbooks, or documenting system architecture.
MCP Servers & Plugins
| Resource |
When to Use |
deploy-on-aws diagram skill |
Generate architecture diagrams for system documentation and runbooks |
deploy-on-aws:awsknowledge |
Reference AWS service docs when writing architecture docs or runbooks — use read_documentation, search_documentation, and recommend to link to official docs rather than paraphrasing |
awslabs.document-loader-mcp-server |
Load external reference documents (PDFs, web pages) as source material for documentation |
aws-serverless plugin |
Use get_lambda_guidance and get_serverless_templates to document serverless patterns, Lambda configurations, and SAM deployment steps |
databases-on-aws plugin |
Use dsql_search_documentation and dsql_recommend to document DSQL schema design, query patterns, and migration procedures |
context7 MCP |
Look up library/framework docs to verify technical accuracy in README examples |
pr-review-toolkit:comment-analyzer plugin |
After writing any documentation — verifies accuracy, staleness risk, and maintainability |
github plugin |
Link to issues, PRs, and discussions from documentation. Create issues for documentation gaps |
README Structure
# Project Name
One-line description of what this does.
## Quick Start
\`\`\`bash
npm install
npm start
\`\`\`
## Configuration
| Variable | Description | Default |
|----------|-------------|---------|
| PORT | Server port | 3000 |
## Usage
[Examples of common operations]
## Development
[How to set up dev environment, run tests]
## License
MIT
API Documentation
Use OpenAPI/Swagger. Minimum per endpoint:
- HTTP method and path
- Request parameters (path, query, body)
- Response codes and schemas
- Authentication requirements
- Example request/response
Runbook Template
# [Service Name] Runbook
## Overview
What this service does, who owns it.
## Architecture
[Use `deploy-on-aws` diagram skill to generate architecture diagram]
## Health Checks
- Endpoint: `GET /health`
- Expected: 200 OK
## Common Issues
### Issue: High latency
**Symptoms**: Response times > 500ms
**Diagnosis**: Check DB connections, cache hit rate
**Resolution**: Scale horizontally, clear cache
## Escalation
- L1: On-call engineer
- L2: Service owner
- L3: Platform team
Architecture Decision Record (ADR)
# ADR-001: Use PostgreSQL for user data
## Status
Accepted
## Context
Need persistent storage for user accounts.
## Decision
Use PostgreSQL on RDS.
## Consequences
- Pro: ACID compliance, familiar tooling
- Con: Operational overhead vs DynamoDB
Spec Artifact Documentation
When documenting within the spec-driven workflow, these artifacts have defined formats (see spec-workflow skill):
| Artifact |
Owner |
Purpose |
spec.md |
fullstack-agent |
Design decisions, constraints, alternatives considered |
design.md |
fullstack-agent |
Architecture, repo structure, infrastructure design |
tasks.md |
fullstack-agent (authored), all teammates (updated) |
Parallelized task groups with completion notes |
review.md |
review-agent |
Severity-rated findings with PASS/FAIL verdict |
sa-review.md |
sa-agent |
Well-Architected findings by pillar, cost estimates |
decisions.md |
any agent via fullstack-agent |
Mid-flight decisions to prevent re-litigation |
When writing documentation for a project that uses the spec workflow, link to relevant specs rather than duplicating their content.
Agent Integration
devops-agent owns READMEs, runbooks, and architecture docs — keeps them next to the code they describe
coding-agent writes inline documentation (function/class/module docs) during implementation
- Both agents delegate to
pr-review-toolkit:comment-analyzer after writing docs to verify accuracy
sa-agent produces architecture review documentation in Well-Architected pillar format, claims and tracks tasks like other teammates
- Use
github plugin to create issues for documentation that needs future updates (e.g., after API changes)
Writing Tips
- Lead with the "what" and "why"
- Use concrete examples over abstract explanations
- Keep it scannable (headers, bullets, tables)
- Update docs when code changes (or automate it)
- Docs are concise and actionable — no filler
- Use the
deploy-on-aws diagram skill for architecture diagrams — don't describe what a diagram can show
- Use
awslabs.document-loader-mcp-server to load external specs or references rather than copy-pasting content
Source: aws-samples/sample-claude-code-agent-team — distributed by TomeVault.
1---2name: documentation-1023description: Technical writing patterns for READMEs, runbooks, and architecture docs, with accuracy verification. Use when writing documentation, creating runbooks, or documenting system architecture. Use when this capability is needed.4---56# Documentation78Use these patterns when writing documentation, creating runbooks, or documenting system architecture.910## MCP Servers & Plugins1112| Resource | When to Use |13|---|---|14| `deploy-on-aws` diagram skill | Generate architecture diagrams for system documentation and runbooks |15| `deploy-on-aws:awsknowledge` | Reference AWS service docs when writing architecture docs or runbooks — use `read_documentation`, `search_documentation`, and `recommend` to link to official docs rather than paraphrasing |16| `awslabs.document-loader-mcp-server` | Load external reference documents (PDFs, web pages) as source material for documentation |17| `aws-serverless` plugin | Use `get_lambda_guidance` and `get_serverless_templates` to document serverless patterns, Lambda configurations, and SAM deployment steps |18| `databases-on-aws` plugin | Use `dsql_search_documentation` and `dsql_recommend` to document DSQL schema design, query patterns, and migration procedures |19| `context7` MCP | Look up library/framework docs to verify technical accuracy in README examples |20| `pr-review-toolkit:comment-analyzer` plugin | After writing any documentation — verifies accuracy, staleness risk, and maintainability |21| `github` plugin | Link to issues, PRs, and discussions from documentation. Create issues for documentation gaps |2223## README Structure2425```markdown26# Project Name2728One-line description of what this does.2930## Quick Start3132\`\`\`bash33npm install34npm start35\`\`\`3637## Configuration3839| Variable | Description | Default |40|----------|-------------|---------|41| PORT | Server port | 3000 |4243## Usage4445[Examples of common operations]4647## Development4849[How to set up dev environment, run tests]5051## License5253MIT54```5556## API Documentation5758Use OpenAPI/Swagger. Minimum per endpoint:59- HTTP method and path60- Request parameters (path, query, body)61- Response codes and schemas62- Authentication requirements63- Example request/response6465## Runbook Template6667```markdown68# [Service Name] Runbook6970## Overview71What this service does, who owns it.7273## Architecture74[Use `deploy-on-aws` diagram skill to generate architecture diagram]7576## Health Checks77- Endpoint: `GET /health`78- Expected: 200 OK7980## Common Issues8182### Issue: High latency83**Symptoms**: Response times > 500ms84**Diagnosis**: Check DB connections, cache hit rate85**Resolution**: Scale horizontally, clear cache8687## Escalation88- L1: On-call engineer89- L2: Service owner90- L3: Platform team91```9293## Architecture Decision Record (ADR)9495```markdown96# ADR-001: Use PostgreSQL for user data9798## Status99Accepted100101## Context102Need persistent storage for user accounts.103104## Decision105Use PostgreSQL on RDS.106107## Consequences108- Pro: ACID compliance, familiar tooling109- Con: Operational overhead vs DynamoDB110```111112## Spec Artifact Documentation113114When documenting within the spec-driven workflow, these artifacts have defined formats (see `spec-workflow` skill):115116| Artifact | Owner | Purpose |117|----------|-------|---------|118| `spec.md` | fullstack-agent | Design decisions, constraints, alternatives considered |119| `design.md` | fullstack-agent | Architecture, repo structure, infrastructure design |120| `tasks.md` | fullstack-agent (authored), all teammates (updated) | Parallelized task groups with completion notes |121| `review.md` | review-agent | Severity-rated findings with PASS/FAIL verdict |122| `sa-review.md` | sa-agent | Well-Architected findings by pillar, cost estimates |123| `decisions.md` | any agent via fullstack-agent | Mid-flight decisions to prevent re-litigation |124125When writing documentation for a project that uses the spec workflow, link to relevant specs rather than duplicating their content.126127## Agent Integration128129- `devops-agent` owns READMEs, runbooks, and architecture docs — keeps them next to the code they describe130- `coding-agent` writes inline documentation (function/class/module docs) during implementation131- Both agents delegate to `pr-review-toolkit:comment-analyzer` after writing docs to verify accuracy132- `sa-agent` produces architecture review documentation in Well-Architected pillar format, claims and tracks tasks like other teammates133- Use `github` plugin to create issues for documentation that needs future updates (e.g., after API changes)134135## Writing Tips136137- Lead with the "what" and "why"138- Use concrete examples over abstract explanations139- Keep it scannable (headers, bullets, tables)140- Update docs when code changes (or automate it)141- Docs are concise and actionable — no filler142- Use the `deploy-on-aws` diagram skill for architecture diagrams — don't describe what a diagram can show143- Use `awslabs.document-loader-mcp-server` to load external specs or references rather than copy-pasting content144145---146> Source: [aws-samples/sample-claude-code-agent-team](https://github.com/aws-samples/sample-claude-code-agent-team) — distributed by [TomeVault](https://tomevault.io).147<!-- tomevault:4.0:skill_md:2026-06-15 -->