AI-native task management for Claude Code. Track tasks in a MASTER_PLAN.md file, pick what to work on, save progress, and ship — all through agent skills.
/master-plan:task → create → /master-plan:next → pick → /master-plan:save → pause → /master-plan:done → ship
Step 1 — Install the plugin (run these two lines inside Claude Code):
/plugin marketplace add endlessblink/master-plan
/plugin install master-plan
Step 2 — Create your first task:
/master-plan:task
That's it. If your project doesn't have a MASTER_PLAN.md yet, the plugin creates one automatically in docs/MASTER_PLAN.md.
Step 3 — Work the loop:
/master-plan:next Pick what to work on
... write code ...
/master-plan:save End of session? Save WIP + push
... next day, different machine ...
/master-plan:next Pick up where you left off
... finish the task ...
/master-plan:done Test + commit + push + mark complete
Already have tasks in a MASTER_PLAN.md? The plugin works with any existing file as long as tasks use
### TASK-123: Title (STATUS)headers. Run/master-plan:nextto start picking from your existing backlog.
- Zero infrastructure — Just a markdown file + git. No Jira, no Linear, no database.
- AI-native — Claude reads the plan, scores priorities, picks tasks, updates status, runs tests, commits. You just approve.
- Session-resilient —
/saveyour work, switch machines, come back tomorrow,/nextpicks up where you left off. - Self-documenting — The MASTER_PLAN.md IS the project history. Nothing to keep in sync.
- Any language, any framework — Works with Node, Python, Rust, Go, or anything else. Auto-detects your test runner.
- Cross-platform — Built on the Agent Skills open standard. Core functionality works in Claude Code, Cursor, GitHub Copilot, Windsurf, and other compatible tools. Interactive task picking is best in Claude Code.
Creates a new task with an auto-generated sequential ID. If no MASTER_PLAN.md exists, creates one from a starter template.
> /master-plan:task
Using task number: 42
? Type: TASK / BUG / FEATURE / INQUIRY
? Priority: P0 (critical) → P3 (backlog)
? Title: Add user authentication
Task added to MASTER_PLAN.md:
- ID: TASK-42
- Title: Add user authentication
- Priority: P1
- Status: PLANNED
Reads your MASTER_PLAN.md, scores tasks by priority and status, shows an interactive picker. Highlights in-progress tasks so you finish what you started.
> /master-plan:next
You have 1 task in progress.
? Which task would you like to work on?
● TASK-42: Add user authentication P1 · PLANNED
BUG-38: Fix login redirect P0 · PLANNED
TASK-40: Refactor API layer P2 · IN PROGRESS
Show more tasks...
> [Selects TASK-42, shows full details, offers to start working]
Filters: /master-plan:next bugs · /master-plan:next planned · /master-plan:next active · /master-plan:next review
Commits and pushes without marking the task done. No tests, no ceremony — just save and go. Perfect for switching machines or ending a session.
> /master-plan:save
## Progress Saved
- Task: TASK-42
- Summary: Added login form and validation
- Commit: a1b2c3d — wip(TASK-42): Added login form and validation
- Status: Still IN PROGRESS
Ready to continue on another machine:
git pull → /master-plan:next
The full completion workflow: run tests, commit, push, and mark the task done in all locations within MASTER_PLAN.md.
> /master-plan:done
## Task Complete
- Task: TASK-42
- Tests: ✅ Passed
- Commit: d4e5f6g — feat(TASK-42): Add user authentication
- Push: ✅ Pushed to origin
- MASTER_PLAN.md: ✅ Updated (table + header + bullets)
Test auto-detection: npm test · pytest · cargo test · go test ./... · make test
Commit format: feat(TASK-XXX): for features, fix(BUG-XXX): for bugs, wip(TASK-XXX): for saves.
Start session
│
▼
/master-plan:next ← Pick a task (scored by priority)
│
▼
Work on it ← Write code, debug, iterate
│
├──► Need to stop? → /master-plan:save (WIP commit + push)
│ Come back later, git pull, /next
│
├──► Found a bug? → /master-plan:task (log it, keep working)
│
└──► Done! → /master-plan:done (test + commit + push + mark complete)
│
▼
/master-plan:next ← Pick the next task
The plugin reads and writes tasks in this format:
### TASK-42: Add user authentication (IN PROGRESS)
**Priority**: P1
**Status**: IN PROGRESS (2026-02-23)
Description of the task...
**Tasks**:
- [ ] Create login form
- [x] Set up auth providerStatuses: PLANNED · IN PROGRESS · REVIEW · PAUSED · ✅ DONE
When a task is marked done, it gets strikethrough on the ID and a status update:
### ~~TASK-42~~: Add user authentication (✅ DONE)Tasks can also appear in a roadmap table — the plugin updates both locations automatically.
Starting fresh? Run
/master-plan:taskand the plugin creates the file for you. A starter template is also available attemplates/MASTER_PLAN.md.
| Prefix | Usage |
|---|---|
TASK-XXX |
Features and improvements |
BUG-XXX |
Bug fixes |
FEATURE-XXX |
Major features |
ROAD-XXX |
Roadmap items |
IDEA-XXX |
Ideas for later |
ISSUE-XXX |
Known issues |
INQUIRY-XXX |
Investigations |
IDs are globally sequential across all types. If TASK-42 and BUG-43 exist, the next ID is 44.
Commit this to your project's .claude/settings.json and teammates are prompted to install automatically:
{
"extraKnownMarketplaces": {
"endlessblink-tools": {
"source": { "source": "github", "repo": "endlessblink/master-plan" }
}
},
"enabledPlugins": {
"master-plan@endlessblink-tools": true
}
}- Claude Code (or any Agent Skills-compatible tool)
- Git
No npm install, no build step, no external dependencies. The skills are pure markdown — Claude does the parsing.
Issues and PRs welcome at github.com/endlessblink/master-plan.
MIT