Task Plan

Task Plan - Generate Task Breakdown from PRD

Analyzes PRD/SPEC files and creates/updates comprehensive task breakdown structures. Supports subcommands: add, status, run.

Subcommands

| Command | Description | |----------------------------------|--------------------------------------| | /task-plan | Analyze PRD, generate/update tasks | | /task-plan docs/PRD.md | Specify PRD path | | /task-plan add "description" | Add a new feature inline | | /task-plan add @docs/spec.md | Add feature from file | | /task-plan status | Show task status table | | /task-plan run | Run all tasks autonomously |

Operating Mode

On each execution:

1. PRD Analysis: Analyze the PRD file from scratch
2. Check Current State: Read existing files in tasks/ directory
3. Detect Changes: Identify new features, changed requirements, or removed sections
4. Incremental Update: Update only changed parts, preserve existing progress

Finding PRD File

1. If provided as argument: $ARGUMENTS
2. Otherwise search project root: PRD.md, SPEC.md, prd.md, spec.md, docs/PRD.md, docs/SPEC.md

Core Steps

Step 1: Find and Read PRD File

Search for PRD/SPEC in project root and docs/, specifications/, specs/ directories.

Step 2: Analyze Existing Task Structure

If tasks/ directory exists:

• Read all feature files (tasks/XXX-*.md)
• Read tasks/tasks-status.md
• Record status of completed and in-progress tasks
• Track highest Feature ID (FXXX) and Task ID (TXXX) for continuation

When PRD is added to an existing project with manually added features, new features must continue from the highest existing IDs. Never reset or conflict.

Step 3: Parse PRD and Compare

Extract: project metadata, feature boundaries, technical stack, user stories, performance criteria, security requirements.

Compare with existing tasks:

• New features: In PRD but not in tasks/
• Changed features: Definition changed in PRD
• Removed features: In tasks/ but not in PRD (warn, don't delete)

Step 4: Create/Update Task Structure

For each new feature, create tasks/XXX-feature-name.md using the [feature template](templates/feature.md).

IMPORTANT RULES:

1. Create 3-6 tasks per feature, each task should be 0.5-3 days of work
2. Tasks should be atomic and independently testable
3. Use realistic effort estimates based on complexity
4. Dependencies MUST reference actual task IDs (T001, T002, etc.) or be "None"
5. Do NOT use vague dependencies like "All backend features" or "Previous tasks"
6. Success criteria must be specific and measurable
7. Technical details should guide implementation
8. Priority levels: P1=Critical, P2=High, P3=Medium, P4=Low

Rules for existing features:

• Preserve completed tasks (don't modify COMPLETED status)
• Preserve in-progress tasks (keep progress)
• Add new tasks for new requirements
• Mark changed requirements with AT_RISK status

Also update:

• [Status tracker](templates/status-tracker.md) at tasks/tasks-status.md
• [Execution plan](templates/execution-plan.md) at tasks/task-execution-plan.md

Step 5: Generate Change Summary

Output: added features, updated features, warnings (removed from PRD), statistics.

Task Properties Standards

Each task must include:

1. Unique ID: TXXX format
2. Status: NOTSTARTED | INPROGRESS | COMPLETED | BLOCKED | AT_RISK | PAUSED
3. Priority: P1 (Critical) | P2 (High) | P3 (Medium) | P4 (Low)
4. Effort: Developer-days (1 day = 6-8 hours)
5. Dependencies: Hard and soft dependencies
6. Success Criteria: At least 3-5 measurable criteria
7. Files to Touch: File paths (new/update/delete)

Task sizing: Atomic tasks 0.5-5 days, Features 1-6 weeks, Milestones 1-3 months.

Subcommand Details

• For /task-plan add: see [subcommands/add.md](subcommands/add.md)
• For /task-plan status: see [subcommands/status.md](subcommands/status.md)
• For /task-plan run: see [subcommands/run.md](subcommands/run.md)

Quality Checklist

• [ ] All tasks have unique IDs
• [ ] Dependencies are not circular
• [ ] Estimates are realistic
• [ ] Success criteria are measurable
• [ ] File paths are correct
• [ ] Critical path is optimized
• [ ] Documentation and test tasks included
• [ ] No mock code or placeholders - all implementations must be production-ready

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: KilimcininKorOglu
• Source: KilimcininKorOglu/cli-tweaks
• License: MIT

Install and usage instructions live in the source repository linked above.