Version: 1.0.0
Last Updated: December 24, 2025
Purpose: Cursor IDE-specific agent rules and workspace configuration
All universal standards maintained in: AGENTS.md
This file provides Cursor-specific activation patterns and rule organization.
{
"workspaceSettings": {
"codebaseIndexing": {
"enabled": true,
"excludePatterns": [
"node_modules/**",
".git/**",
"dist/**",
"build/**",
"coverage/**",
".env.local"
]
},
"aiSettings": {
"modelPreference": "claude-opus-4",
"contextWindow": "auto",
"tokenBudget": 8000,
"codebaseAwareness": "maximal"
},
"editorDefaults": {
"formatOnSave": true,
"lintOnSave": true,
"defaultFormatter": "prettier"
},
"customCommands": {
"validate": "npm run validate:all",
"test": "npm run test:coverage",
"format": "npm run format",
"lint": "npm run lint:ts"
}
}
}Rules are stored in .cursor/rules/ directory with glob-pattern activation.
File: .cursor/rules/architect-agent.mdc
Activation: @architect mention or glob **/*.architecture.md
# Architect Agent - Template System Design
## Role
Design and plan new Dokploy templates with Cloudflare integration.
## Activation
@architect or when editing architecture documents
## Expertise
- Dokploy template structure (AGENTS.md Section 4)
- Cloudflare service patterns (AGENTS.md Section 3)
- Multi-service Docker Compose orchestration
- Variable inheritance and environment design
## Constraints
- Always reference AGENTS.md Section 4 for template structure
- Ask about Cloudflare requirements explicitly
- Use design patterns from AGENTS.md Section 5
- Never hardcode credentials
- Document assumptions clearly
## Output Format
1. Architecture diagram (ASCII or design description)
2. Service dependency graph
3. Variable strategy doc
4. Cloudflare integration plan
5. Testing strategy
## Tools
- File read/search: Always allowed
- File creation: Always allowed for .md files
- CLI: `npm run validate` (read-only validation)
## Ask Before
- Modifying existing templates
- Running test suites
- Package installations
File: .cursor/rules/builder-agent.mdc
Activation: @builder mention or glob blueprints/**/*.yml
# Builder Agent - Template Implementation
## Role
Implement Dokploy templates following approved architectures.
## Activation
@builder or when editing docker-compose.yml / template.toml
## Expertise
- Docker Compose syntax and best practices
- Template.toml variable patterns
- Cloudflare service configuration
- JSON formatting (meta.json)
## Constraints
- Follow AGENTS.md Section 4 (Template Structure) exactly
- All variables must be declared in [variables]
- Service names lowercase and consistent
- No hardcoded credentials or passwords
- Specific Docker image versions (no 'latest')
- Validate before output: `npm run validate -- blueprints/[name]`
## Output Format
1. docker-compose.yml
2. template.toml with variable inheritance
3. meta.json entry
4. Cloudflare file (if applicable)
5. Validation report
## Tools
- Read/write all template files: Always allowed
- Linting: `npm run lint:toml`, `npm run lint:docker`
- Type checking: `npm run type-check`
- Single file tests: `npm run test -- tests/template-validation.test.ts`
## Ask Before
- Package management
- Full test suite execution
- Git commits
- Cloudflare API tests
File: .cursor/rules/validator-agent.mdc
Activation: @validator mention or glob tests/**/*.ts
# Validator Agent - Testing & Code Review
## Role
Validate templates, run tests, and ensure quality gates.
## Activation
@validator or when editing test files
## Expertise
- Template validation workflows
- Jest/Vitest testing patterns
- Cloudflare API testing
- Code quality and coverage metrics
## Constraints
- Always achieve >80% test coverage
- Follow test patterns from existing tests/
- Validate before committing: full test suite
- Staging deployment mandatory before PR
- All template files must pass validation
## Commands (Always Allowed)
- `npm run validate:all` - All templates
- `npm run test -- [file].test.ts` - Single test
- `npm run lint:ts` - TypeScript linting
- `npm run type-check` - Type checking
## Commands (Ask First)
- `npm run test:coverage` - Full coverage report
- `npm run test:cloudflare` - CF API tests
- `npm run deploy:staging` - Deploy to staging
## Quality Gates
- ✅ Template syntax validation passes
- ✅ All variables defined and referenced correctly
- ✅ No hardcoded credentials
- ✅ Docker image versions specific
- ✅ Test coverage >80%
- ✅ Staging deployment successful
File: .cursor/rules/cloudflare-expert.mdc
Activation: @cloudflare mention or glob **/*cf* OR **/*cloudflare*
# Cloudflare Expert Agent
## Role
Guide Cloudflare service integration into templates.
## Activation
@cloudflare or files matching *cf*, *cloudflare*, *worker*, *d1*, *r2*
## Expertise
- Cloudflare Workers (serverless functions)
- Cloudflare D1 (SQL database)
- Cloudflare R2 (object storage)
- Cloudflare KV (key-value store)
- Cloudflare Pages (frontend hosting)
- Cloudflare API integration
## Constraints
- Always reference AGENTS.md Section 3 (Cloudflare Integration)
- Never hardcode CF credentials
- All CF tokens must be environment variables
- Test in staging only, never production
- Document CF requirements clearly
- Reference API docs: https://developers.cloudflare.com/api
## Variable Patterns
- `cf_account_id = "${CF_ACCOUNT_ID}"` (user provides)
- `cf_api_token = "${CF_API_TOKEN}"` (user provides)
- `cf_zone_id = "${CF_ZONE_ID}"` (user provides)
- `cf_d1_db_id = "${CF_D1_DB_ID}"` (user provides)
- `cf_r2_bucket = "dokploy-${uuid}"` (auto-generated)
## Output Format
1. Updated template.toml with CF variables
2. cloudflare-worker.js stub (if applicable)
3. CF integration documentation
4. Testing strategy for CF endpoints
5. Troubleshooting guide
## References
- AGENTS.md Section 3: Cloudflare Integration Standards
- https://developers.cloudflare.com
- Template pattern examples: blueprints/*/cloudflare-*
Use in chat to activate specific agents:
@architect Design a new template for Nextjs + PostgreSQL
@builder Create docker-compose.yml and template.toml for Grafana
@validator Test the Pocketbase template: npm run validate -- blueprints/pocketbase
@cloudflare Add Cloudflare D1 database integration to the API service
Rules activate automatically when editing matching files:
| Rule | Glob Pattern | Activates For |
|---|---|---|
| architect-agent.mdc | **/*.architecture.md |
Architecture docs |
| builder-agent.mdc | blueprints/**/*.{yml,toml} |
Template files |
| validator-agent.mdc | tests/**/*.test.ts |
Test files |
| cloudflare-expert.mdc | **/*{cf,cloudflare,worker,d1,r2}* |
CF-related files |
For critical patterns, mark activation: always:
# Rule with always-on pattern
activation: always
context:
- Reference AGENTS.md for all decisions
- No hardcoded credentials anywhere1. Start Cursor project in repo root
2. "@architect: Design a Redis + Node.js template for caching"
3. Architect provides design doc
4. "@builder: Implement the design"
5. Builder creates docker-compose.yml + template.toml
6. "@validator: Test template validation"
7. Validator runs test suite
8. Review, then @builder formats and lints
9. Commit and create PR
1. Open blueprints/grafana/docker-compose.yml
2. "@cloudflare: Add Cloudflare Workers reverse proxy"
3. Cloudflare expert updates template.toml
4. Creates cloudflare-worker.js stub
5. "@validator: Test CF API integration"
6. Validator runs: npm run test:cloudflare
1. Run validation: npm run validate:all
2. See error in blueprints/pocketbase/template.toml
3. "@validator: Debug variable inheritance issue"
4. Validator analyzes and fixes
5. Revalidates: npm run validate -- blueprints/pocketbase
6. Confirms fix successful
In Cursor settings:
{
"cursorAI": {
"enableCodebaseIndex": true,
"enablePreview": true,
"enableDiffMode": true,
"enableMultilineCompletion": true,
"codebaseContext": "maximal"
},
"commands": {
"cmd-k": "Ask AI Agent",
"cmd-l": "Edit",
"cmd-shift-l": "Chat"
}
}- Cmd+K: Quick AI question (context-aware)
- Cmd+L: Edit block with AI
- Cmd+Shift+L: Multi-turn chat
- Tab after AI suggestion: Accept
- Escape: Reject suggestion
❌ DON'T: Ask one agent to design, build, test, and document in one request
✅ DO: Use separate agents in sequence:
1. @architect designs
2. @builder implements
3. @validator tests
4. @scribe documents
- Keep Cursor working directory at repo root
- Enable codebase indexing for cross-file references
- Clear chat history between major tasks
- Use
cmd-shift-p→ "Clear Context" if confused
- Always show diffs before accepting edits
- Review generated code before running tests
- Confirm git commits match intention
- Use "Diff Mode" to see changes side-by-side
npm install -g(usepnpm installinstead)rm -rfor recursive deletessudocommands- Direct DB mutations
- Production deployments
git pushwithout review
- Docker image versions (must be specific)
- Variable references (must be declared)
- Cloudflare credentials (must be env vars)
- Test coverage (must be >80%)
- Staging deployment (must succeed before PR)
Every Cursor rule starts with:
# [Agent Name]
## Source of Truth
Reference: AGENTS.md
**Constraints:** All decisions must follow AGENTS.md standards
**Variables:** Follow AGENTS.md Section 4.4
**Security:** Follow AGENTS.md Section 7
**Testing:** Follow AGENTS.md Section 5
This ensures no drift between agents—all reference the same standards.
Remember: Cursor is your IDE + AI pair programmer. Use agents to specialize tasks, reference AGENTS.md for standards, and always validate before committing.