task-management

name	task-management
description	Create, update, list, prioritize, and triage tasks. TRIGGER when: user says "add task", "todo", "what are my tasks", "mark done", "triage", "prioritize", or manages any work item tracking.
argument-hint	[add|list|done|triage|delete] [task description or ID]
user-invocable	true

Task Management

A structured, file-based task tracking system. Each task is an individual markdown file with YAML frontmatter, indexed in a central manifest.

First-Run Setup

Before any operation, ensure the task system exists:

1. Check if ~/.claude/tasks/ exists — if not, create with mkdir -p ~/.claude/tasks/
2. Check if ~/.claude/tasks/TASKS.md exists — if not, create:

# Task Index

<!-- Auto-maintained. Do not edit manually. -->

## In Progress

## To Do

## Blocked

## Done

Task Schema

Each task is a file at ~/.claude/tasks/TASK-{NNN}.md:

---
id: TASK-001
title: "Short imperative description (max 80 chars)"
status: todo | in-progress | done | blocked
priority: critical | high | medium | low
created: 2026-04-08
updated: 2026-04-08
due: 2026-04-15          # optional — ISO date
completed: null           # set to ISO date when done
tags: [feature, backend]  # optional — freeform labels
blocked-by: TASK-003      # optional — dependency
source: "gh#45"           # optional — link to GitHub issue, PR, or external ref
---

Detailed description of what needs to be done and why.

## Acceptance Criteria
- [ ] Criterion 1
- [ ] Criterion 2

## Notes
- YYYY-MM-DD: Progress note or context update

Operations

Add — $ARGUMENTS starts with "add" or user describes a new task

1. Parse input: Extract title, priority, due date, tags from the user's description. Infer what you can; ask only for what's ambiguous.
2. Generate ID: Read existing task files to find the highest ID, increment by 1. Pad to 3 digits (e.g., TASK-001).
3. Normalize dates: Convert relative dates to absolute ISO format. Use today's date as reference.
4. Default values: status: todo, priority: medium, created: today, updated: today.
5. Write task file: Create ~/.claude/tasks/TASK-{NNN}.md with full schema.
6. Update index: Add entry to the appropriate section in TASKS.md: - [TASK-{NNN}](TASK-{NNN}.md) — {title} ({priority}, due {date})
7. Confirm: Show the created task with its ID.

List — $ARGUMENTS starts with "list" or user asks to see tasks

1. Read TASKS.md and present grouped by status.
2. Filtering: Support filters from $ARGUMENTS:
   • list high — filter by priority
   • list blocked — filter by status
   • list backend — filter by tag
   • list overdue — due date < today and status != done
3. Sorting: Within each status group, sort by: priority (critical > high > medium > low), then due date (soonest first), then creation date.
4. Output format:

## Tasks (5 total: 1 in-progress, 3 todo, 1 blocked)

### In Progress
- TASK-005: Migrate user table (high, due tomorrow)

### To Do
- TASK-007: Write API docs (medium, no due date)
- TASK-008: Add rate limiting (medium, due 2026-04-20)
- TASK-009: Update onboarding flow (low)

### Blocked
- TASK-006: Deploy auth service — blocked by TASK-005

Done — $ARGUMENTS starts with "done" or user marks a task complete

1. Find the task by ID or keyword match. If ambiguous, present options and ask.
2. Update the task file:
   • Set status: done
   • Set completed: {today}
   • Set updated: {today}
3. Move the entry in TASKS.md from its current section to ## Done.
4. If the completed task was blocking other tasks (blocked-by references), notify the user and offer to unblock them.
5. Confirm completion.

Update — user wants to change a task's priority, due date, description, or status

1. Find the task by ID or keyword.
2. Apply the requested changes to the task file frontmatter and/or body.
3. Always set updated: {today}.
4. Add a timestamped note under ## Notes describing the change.
5. Update TASKS.md if the status, priority, or title changed.
6. Confirm what was updated.

Triage — $ARGUMENTS starts with "triage" or user asks to review tasks

Triage walks through tasks that need attention:

1. Identify candidates:
   • Overdue: due < today and status != done
   • Stale: updated > 7 days ago and status is todo or in-progress
   • Unestimated: no due date and priority is high or critical
   • Long-running: in-progress for > 14 days
2. Present each with context and ask the user to choose an action:
   • Reprioritize: change priority
   • Reschedule: set or update due date
   • Close: mark as done or remove
   • Skip: leave as-is for now
3. Apply changes as the user decides.
4. Summary: After triage, show counts of actions taken.

Delete — $ARGUMENTS starts with "delete" or "remove"

1. Find the task by ID or keyword.
2. Require explicit confirmation — show the full task before deleting.
3. Remove the task file.
4. Remove the entry from TASKS.md.
5. Check if any other tasks have blocked-by referencing this task. If so, warn the user.
6. Confirm deletion.

Integrity Rules

• Single source of truth: Task files are authoritative. TASKS.md is a derived index.
• Index repair: If TASKS.md is out of sync with task files (missing entries, wrong status), rebuild it from the task files.
• No duplicates: Before creating, search existing tasks for similar titles. Warn if a near-duplicate exists.
• ID immutability: Once assigned, a task ID never changes. Deleted IDs are not reused.
• Audit trail: Every status change gets a timestamped note in the task file.
• Atomic updates: Always update both the task file AND TASKS.md in the same operation. Never leave them inconsistent.

Quality Checklist

• Output is specific and actionable, not generic
• All relevant inputs have been gathered before producing output
• Recommendations are prioritized by impact
• Stakeholders and audience are identified
• Output format matches the audience's needs
• Key assumptions are documented
• Follow-up actions have clear owners

Edge Cases

• If input data is incomplete, state assumptions explicitly and flag gaps
• For time-sensitive situations, prioritize speed over comprehensiveness
• If multiple stakeholders have conflicting needs, document the tradeoffs
• For first-time use, start with a simplified version and iterate
• Adapt the depth and detail to the audience's expertise level