rbarcante/claude-conductor/conductor-methodology
Claude Code and agent skill by rbarcante/claude-conductor
- community
Checking that the SKILL.md file is still in the repository.
Use this skill when the user asks about Conductor concepts, workflow patterns, track lifecycle, plan structures, or context-driven development principles. Also activate when working with conductor/ directory files or when implementing tasks following TDD and verification protocols.
Install
npx vskill@latest install rbarcante/claude-conductor/conductor-methodologySKILL.md
Conductor Methodology Skill
Conductor is a context-driven development framework that structures software development through a rigorous lifecycle: Context → Spec & Plan → Implement. This skill provides guidance on Conductor's principles, file structures, and workflows.
Core Philosophy
Measure twice, code once. Conductor treats context as a managed artifact alongside code, transforming repositories into a single source of truth that drives every agent interaction with deep, persistent project awareness.
Key Principles
- Plan before you build: Create specs and plans that guide implementation
- Maintain context: Ensure AI follows style guides, tech stack choices, and product goals
- Iterate safely: Review plans before code is written
- Work as a team: Set project-level context shared across the team
Project Structure
Core Context Files (conductor/)
Located in the conductor/ directory at project root:
- `product.md`: Product vision, users, goals, and features
- `product-guidelines.md`: Prose style, brand messaging, visual identity
- `tech-stack.md`: Technology choices, languages, frameworks, databases
- `workflow.md`: Development workflow (TDD, commit strategy, coverage requirements)
- `code_styleguides/`: Language-specific coding standards
- `tracks.md`: Master list of all tracks with status and links
Track Structure (conductor/tracks/)
Each track lives in conductor/tracks/<track_id>/:
- `spec.md`: Detailed requirements for the track
- `plan.md`: Hierarchical plan with phases, tasks, and sub-tasks
- `decisions.md`: Architecture Decision Records for track decisions
- `review.md`: Auto-generated code review report on track completion
- `metadata.json`: Track metadata (ID, type, status, timestamps)
Track Lifecycle
Tracks represent high-level units of work (features, bugs, chores):
- Creation (
/conductor:newTrack): Generate spec and plan - In Progress: Mark as
[~]when work begins - Completed: Mark as
[x]when all tasks finish - Cleanup: Archive or delete completed tracks
Plan File Structure
Plans use hierarchical markdown with status markers:
Status Markers
[ ]- Pending[~]- In progress[x]- Completed
Plan Format Rules
- Phases: Top-level markdown headers (
# Phase Name) - Parent Tasks: Bullet points with
- [ ] Task: Description - Sub-tasks: Indented bullet points with
- [ ] Description - Commit SHAs: Appended after completion (e.g.,
- [x] Task: Description [abc1234]) - Phase Checkpoints: Appended to phase headers (e.g.,
# Phase 1 [checkpoint: def5678])
Workflow Patterns
Test-Driven Development (TDD)
Standard workflow from workflow.md:
- Select Task: Choose next pending task from plan
- Mark In Progress: Change
[ ]to[~]in plan - Write Failing Tests (Red): Create tests that define expected behavior
- Implement to Pass (Green): Write minimal code to pass tests
- Refactor: Improve code while keeping tests passing
- Verify Coverage: Ensure >80% coverage (or configured percentage)
- Document Deviations: Update tech-stack.md if needed
- Update Plan Status: Mark task as
[x]in plan.md (pre-commit) - Confirm Commit: Ask user to commit or skip
- Commit Code + Plan: Stage code changes and plan.md together, commit with descriptive message
- Attach Task Summary: Use git notes for detailed summary (optional)
- Record SHA in Plan: Append commit SHA to task line on disk (included in next commit)
Phase Completion Protocol
When a phase completes:
- Ensure Test Coverage: Verify all changed files have tests
- Execute Automated Tests: Run full test suite with
CI=true - Propose Manual Verification Plan: Provide step-by-step verification steps
- Await User Feedback: Get explicit confirmation
- Record Phase Reference: Get last task commit SHA, append
[checkpoint: <sha>]to phase header in plan (on disk only) - Announce Completion: Inform user the phase is complete and verified
Commit Message Patterns
- Implementation:
feat(module): Add feature description - Bug Fix:
fix(module): Fix issue description - Tests:
test(module): Add tests for feature - Refactoring:
refactor(module): Refactor description - Documentation:
docs(module): Update documentation - Chores:
chore: Maintenance task description
Note: Plan.md updates are bundled into code commits — no separate conductor-specific commits are created. Track creation uses standard commit types (e.g., chore: Create track 'description').
Git Notes Usage
Conductor uses git notes for auditable tracking:
Task Notes
Attached to implementation commits with format:
Phase Verification Notes
Optionally attached to the last task commit of a phase:
Question Types for Interactive Workflows
When gathering requirements or context:
Additive Questions
For brainstorming and scope (users, goals, features):
- Allow multiple answers
- Add "(Select all that apply)"
- Present as options A, B, C, D, E
Exclusive Choice Questions
For singular commitments (technology selection, workflow rules):
- Require single answer
- Do NOT add "(Select all that apply)"
- Present as options A, B, C, D, E
Option Format
Always include:
- A, B, C: Suggested answers
- D: "Type your own answer"
- E: "Autogenerate and review <filename>"
Track ID Format
Track IDs follow the pattern: shortname_YYYYMMDD
Examples:
darkmode_20260113- Add dark mode featurebugfix_20260113- Fix login issuerefactor_20260113- Refactor database layer
State Management
Setup State (conductor/setup_state.json)
Tracks setup progress for resumability:
Possible values:
""- Initial state"2.1_product_guide"- Product guide complete"2.2_product_guidelines"- Guidelines complete"2.3_tech_stack"- Tech stack complete"2.4_code_styleguides"- Style guides copied"2.5_workflow"- Workflow configured"3.3_initial_track_generated"- Setup fully complete
Track Metadata (metadata.json)
Best Practices
When Creating Specs
- Ask context-aware questions based on product.md and tech-stack.md
- Provide 2-3 plausible options for each question
- Ask questions sequentially (one at a time)
- Include acceptance criteria and out-of-scope items
- Get user confirmation before proceeding
When Creating Plans
- Read and follow the workflow.md methodology
- Break features into testable tasks
- Include phase completion verification tasks
- Use hierarchical structure (phases > tasks > sub-tasks)
- Add status markers to EVERY item
- Get user confirmation before starting work
When Implementing
- Follow the workflow.md task lifecycle precisely
- Never skip test-writing step in TDD
- Verify tool call success before proceeding
- Update plan.md status in real-time
- Use git notes for audit trails
- Get user confirmation for phase completions
When Syncing Documentation
- Only update docs when track completes (
[x]) - Analyze spec.md for significant changes
- Propose changes with diff format
- Get explicit user approval before editing
- Be very conservative with product-guidelines.md
- Commit all doc changes together
Reverting Work
Conductor provides git-aware revert for logical units:
- Identify Target: Track, phase, or task
- Find Commits: Implementation + plan updates + track creation (if applicable)
- Handle Rewritten History: Search for similar commit messages if SHA missing
- Present Plan: Show all commits to be reverted
- Get Confirmation: Explicit user approval required
- Execute: Revert in reverse chronological order
Integration with Claude Code
All commands are invoked with /conductor:<command>:
/conductor:setup- Initialize project/conductor:newTrack [description]- Create new track/conductor:implement [track]- Execute track plan/conductor:status- Show progress overview/conductor:revert [target]- Revert work
Commands use Claude Code tools (Read, Write, Edit, Bash, Glob, Grep) and follow plugin best practices.
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/conductor-methodology/SKILL.md
Popularity
- Repository stars
- 43
- Repository forks
- 1
- 7-day trend score
- 19
Stars and forks are for the whole GitHub repository, not this skill alone. Refreshed 24 Mar 2026, 7 months ago.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/rbarcante/claude-conductor/conductor-methodology)