Workflow Orchestrator
Always-on context detection and skill chaining for Claude Code projects.
How It Works
This skill runs in the background, detecting context and activating appropriate skills:
| Context Detected | Skill Activated | Action |
|---|---|---|
| New project creation | project-init | Scaffold standard structure |
| Task completion | doc-maintenance | Update PLAN.md, README.md |
| Bug fix in tool | tdd + no-workarounds | TDD workflow, no manual workarounds |
| Feature completion | dogfood-skills | Run skills scan, install recommendations |
| Non-obvious discovery | claudeception | Extract as new skill |
| Parallel tasks needed | agent-orchestration | Coordinate agents via AGENTS.md |
| Research needed | research-to-plan | Web search, write RESEARCH.md |
| Session end | repo-hygiene | Clean test-skill-* artifacts |
Conditional Chaining
Skills chain based on context:
New Project Flow
project-init → dogfood-skills → tdd (ready)
Task Completion Flow
task complete → doc-maintenance → repo-hygiene
Bug Fix Flow
bug detected → tdd (RED) → tdd (GREEN) → tdd (REFACTOR) → doc-maintenance
Feature Completion Flow
feature done → dogfood-skills → repo-hygiene → doc-maintenance
Feature Development Flow
research-to-plan → tdd → doc-maintenance → dogfood-skills → repo-hygiene
Testing Pipeline
tdd → suggest-tests → unit-test-workflow → property-based-testing → repo-hygiene
Context Detection
The orchestrator detects context from:
- User commands: "create project", "fix bug", "add feature"
- File changes: New files, modified tests, config changes
- Git state: Branch names, commit messages
- Error messages: Test failures, build errors
- Conversation history: Previous actions and outcomes
Skill Relationships
| Skill | Triggers | Chains To |
|---|---|---|
| project-init | "create", "scaffold", "initialize" | dogfood-skills |
| doc-maintenance | Task completion | repo-hygiene |
| agent-orchestration | Parallel tasks | doc-maintenance |
| gitignore-hygiene | After commits | repo-hygiene |
| research-to-plan | Research needed | tdd |
| tdd | Bug fix, new feature | suggest-tests, doc-maintenance |
| suggest-tests | After tdd | unit-test-workflow |
| unit-test-workflow | Test generation | property-based-testing |
| property-based-testing | Serialization patterns | repo-hygiene |
| repo-hygiene | Session end, cleanup | doc-maintenance (feature flows) |
Standard Project Structure
project/
├── CLAUDE.md # Project guidance for Claude
├── README.md # User-facing documentation
├── PLAN.md # Remaining work tracker
├── RESEARCH.md # Investigation notes
├── AGENTS.md # Agent coordination
├── .gitignore # Comprehensive patterns
└── .claude/
└── skills/ # Project-specific skills
When Skills Run
Automatically (Context Detection)
- project-init: User mentions creating a project
- doc-maintenance: After any task marked complete
- repo-hygiene: End of session or before commit
On Explicit Invocation
/project-init [name]: Create new project/doc-maintenance: Update documentation/agent-orchestration [task]: Coordinate agents/gitignore-hygiene: Clean gitignore/research-to-plan [topic]: Research to plan
Chained (Triggered by Other Skills)
- tdd triggers after bug fixes
- dogfood-skills triggers after feature completion
- repo-hygiene triggers as terminal skill
Rationalizations (Do Not Skip)
| Rationalization | Why It's Wrong | Required Action |
|---|---|---|
| "I'll update docs later" | Later never comes | doc-maintenance now |
| "Small change, no docs" | Small changes accumulate | Update PLAN.md |
| "Agents are overkill" | Parallel work saves time | Consider orchestration |
| "Gitignore is fine" | Hidden files get committed | Run hygiene check |
| "I know the research" | Knowledge cutoff may be stale | Web search for 2026 |
| "Skip the tests" | Bugs escape | TDD required |
| "Skip cleanup" | Slop accumulates | repo-hygiene at end |
Integration
This orchestrator connects all workflow skills:
- project-init: Scaffolds new projects
- doc-maintenance: Updates documentation
- agent-orchestration: Coordinates parallel work
- gitignore-hygiene: Maintains gitignore
- research-to-plan: Converts research to plans
- tdd: Test-driven development
- no-workarounds: Prevents manual workarounds
- dogfood-skills: Enforces dogfooding
- repo-hygiene: Final cleanup
References
For detailed guidance, see:
- file-structure.md - Standard project layouts