Manages a multi-agent team to turn design docs into structured OpenSpec task lists.

Install

mkdir -p .claude/skills/plan-kurianoff && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/18170" && unzip -o skill.zip -d .claude/skills/plan-kurianoff && rm skill.zip

Installs to .claude/skills/plan-kurianoff

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Run Phase 2 (Planning) of the privacyclaw feature workflow. Coordinates PM, Architect, Investigator, and Contrarian to produce a validated OpenSpec task list from a Design Document. Can be invoked standalone by passing a Design Phase Handoff as the argument.
258 charsno explicit “when” triggerlonger than Claude Code's old 250-char listing cap (fine on current versions)
Advanced

Key capabilities

  • Coordinate PM, Architect, Investigator, and Contrarian agents
  • Generate a validated OpenSpec task list from a Design Document
  • Ensure all operations occur within an isolated worktree
  • Facilitate agent communication using Agent Handoffs
  • Validate the final task list using `openspec validate`
  • Produce a Phase Handoff detailing completion status and artifacts

How it works

This skill coordinates multiple agents (PM, Architect, Investigator, Contrarian) within an isolated worktree to transform a Design Document into a validated OpenSpec task list. It uses a structured Agent Handoff format for communication and performs validation before producing a final Phase Handoff.

Inputs & outputs

You give it
A Design Phase Handoff document
You get back
A validated OpenSpec task list and a Phase Handoff

When to use plan

  • Planning feature implementation
  • Turning design docs into tasks
  • Coordinating architectural review

About this skill

Phase 2 — Planning

You are the Phase 2 coordinator. Your job is to turn the Design Document into a complete, validated OpenSpec task list that the Development phase can execute without ambiguity.

Input: $ARGUMENTS

Extract from the input:

  • feature: the feature description
  • branch: the feature branch (feature/<slug>)
  • design_doc: path to the Design Document (from the Design handoff Artifacts)
  • decisions: key design decisions from the Design handoff
  • for_next: context the Planning phase needs from Design
  • WORKTREE: the absolute path to the isolated workflow worktree (required; if missing, derive as $(git rev-parse --show-toplevel)/../worktrees/<branch-slug>)

Working directory

All operations in this phase must happen inside <WORKTREE>, never in the main repository working tree.

Rules that apply to this coordinator and to every agent it invokes:

  • File reads/writes: use the absolute path <WORKTREE>/<relative-path>
  • openspec commands: cd "<WORKTREE>" && openspec <command>
  • Every agent message must include WORKTREE: <worktree_path> so agents apply the same rule without ambiguity.

Agent coordination protocol

Try team-based coordination first:

TeamCreate({ name: "plan-team", agents: ["pm", "architect", "investigator", "contrarian"] })
SendMessage({ to: "pm", message: "<task + context>" })

Use SendMessage to pass Agent Handoffs between agents. Fall back to sequential Agent tool calls if TeamCreate fails. Do not retry teams more than once.


Agent Handoff format

Same format used in all phases:

--- AGENT HANDOFF ---
From:     <agent name>
To:       <next agent>
Status:   complete | blocked
Done:
  - <key action taken>
Decisions:
  - <decision + rationale, or "none">
Findings:
  - <finding + severity, or "none">
Open:
  - <item + owner, or "none">
Pass forward:
  <2–3 sentences of critical context for the next agent>
--- END HANDOFF ---

Workflow

Step 1 — PM + Architect: joint planning kickoff

Invoke pm and architect together to study the Design Document and agree on the implementation approach before any task list is drafted. Task:

PM: read the Design Document at <design_doc> and the design decisions from the handoff. Identify the major implementation areas and likely task groupings.

Architect: review the same document. For each area PM identifies, assess: what existing code is touched, what the natural implementation order is, and whether any areas carry hidden complexity that affects scope.

Together, agree on: the implementation areas, their order, and any scope boundaries. Produce a joint Agent Handoff that documents the agreed approach. This becomes the blueprint PM uses to draft tasks.

Step 2 — PM: create OpenSpec proposal

Invoke pm with the joint kickoff handoff and Design Document. Task:

Working directory: <WORKTREE> — read and write files only within this directory. Do not access the main repository working tree. WORKTREE: <worktree_path>

Using the agreed approach from the planning kickoff, run cd "<WORKTREE>" && openspec proposal to scaffold a new change. Create:

  • <WORKTREE>/openspec/changes/<id>/proposal.md
  • <WORKTREE>/openspec/changes/<id>/design.md (if the design warrants it)
  • <WORKTREE>/openspec/changes/<id>/tasks.md — first draft task list

Tasks must be small, ordered, and independently verifiable. Each task must include: what to implement, how to verify it is done. Produce an Agent Handoff with the OpenSpec change id and task count.

Step 2 — Architect: review task list

Invoke architect with the PM handoff and task list path. Task:

Working directory: <WORKTREE> — read and write files only within this directory. Do not access the main repository working tree. WORKTREE: <worktree_path>

Review <WORKTREE>/openspec/changes/<id>/tasks.md. Flag:

  • Missing tasks (things the Design requires that have no task)
  • Wrong sequencing (dependencies not respected)
  • Under-specified tasks (not enough detail to implement without guessing)
  • Tasks that are too large (should be split)

Produce an Agent Handoff listing every issue found.

Step 3 — PM ↔ Architect cycle

Pass the Architect handoff back to pm. PM updates the task list to address every Architect finding. Then invoke Architect again to re-review.

Repeat until the Architect's handoff contains no open items.

PM has final say on task structure. If PM and Architect disagree on how to structure a task, PM's decision stands unless Architect raises a dependency or correctness concern (not a preference).

Step 4 — Investigator and Contrarian: independent review

Invoke investigator and contrarian independently (in parallel if teams are available) against the current task list. Each produces an Agent Handoff.

Investigator task:

Working directory: <WORKTREE> — search and read files only within this directory. Do not access the main repository working tree. WORKTREE: <worktree_path>

Read <WORKTREE>/openspec/changes/<id>/tasks.md. Identify tasks that:

  • Touch code paths with hidden complexity not reflected in the task scope
  • Have implicit dependencies on other tasks not marked
  • Make assumptions about current behaviour that may be wrong Produce an Agent Handoff with findings. Do not propose fixes.

Contrarian task:

Working directory: <WORKTREE> — read files only within this directory. Do not access the main repository working tree. WORKTREE: <worktree_path>

Read <WORKTREE>/openspec/changes/<id>/tasks.md and the Design Document. Challenge:

  • Tasks that are too optimistic about implementation effort
  • Missing tasks for error handling, edge cases, or rollback
  • The overall sequencing — is there a better order?
  • Anything that would cause Phase 3 to stall mid-way Produce an Agent Handoff.

Step 5 — Architect: filter and route feedback

Pass both the Investigator and Contrarian handoffs to architect. Task:

For each finding from Investigator and Contrarian: decide whether it requires a task change. Route actionable findings to PM. Dismiss non-actionable items with a rationale. Produce an Agent Handoff with routing decisions.

Pass Architect's routing decisions to pm. PM creates, updates, or deletes tasks accordingly and produces a final Agent Handoff.

Step 6 — User escalation

If scope, priority, or edge-case decisions arise that require user input:

  1. Collect all pending questions.
  2. Return a Phase Handoff with Status: blocked and questions in Open.

Step 7 — Validate

Run:

cd "<WORKTREE>" && openspec validate <id> --strict

Resolve every issue reported before proceeding. If validation cannot be made to pass, return a blocked Phase Handoff with the validation errors in Open.


Team cleanup

If TeamCreate succeeded earlier, shut down all agents and delete the team before producing the Phase Handoff:

SendMessage({ to: "pm",           message: {type: "shutdown_request"} })
SendMessage({ to: "architect",    message: {type: "shutdown_request"} })
SendMessage({ to: "investigator", message: {type: "shutdown_request"} })
SendMessage({ to: "contrarian",   message: {type: "shutdown_request"} })
TeamDelete()

If TeamCreate was never called (sequential fallback path), skip this section.


Phase completion

Phase 2 is complete when:

  • openspec validate <id> --strict passes with no issues
  • Architect has approved the final task list (no open items in last Architect handoff)
  • All user questions have been answered (or none were raised)

Produce a Phase Handoff:

=== PHASE HANDOFF ===
Phase:     Planning
Status:    complete  (or: blocked — <reason>)
Feature:   <feature description>
Branch:    <branch>
Artifacts:
  <WORKTREE>/openspec/changes/<id>/proposal.md
  <WORKTREE>/openspec/changes/<id>/tasks.md
  <WORKTREE>/openspec/changes/<id>/design.md  (if created)
Decisions: <bullet list of key planning decisions>
For next:  <what Development needs: OpenSpec change id, task count,
            any constraints or known risks Development must plan around>
Open:      <user questions still pending, or "none">
=== END HANDOFF ===

When not to use it

  • When the input is not a Design Phase Handoff
  • When the project does not use OpenSpec for task management
  • When an isolated worktree cannot be created or used

Limitations

  • Requires a Design Phase Handoff as input
  • All operations must happen inside a specified worktree
  • Relies on `openspec` commands for task management

How it compares

This skill orchestrates a multi-agent workflow for planning and task list generation, which is different from a manual process of reviewing design documents and creating tasks.

Compared to similar skills

plan side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
plan (this skill)03moReviewAdvanced
create-plan367moReviewBeginner
project-planner329moReviewIntermediate
system-design198moNo flagsIntermediate

Try saying

Example prompts that trigger this skill in your AI assistant.

You might also like

create-plan

antinomyhq

Generate detailed implementation plans for complex tasks. Creates comprehensive strategic plans in Markdown format with objectives, step-by-step implementation tasks using checkbox format, verification criteria, risk assessments, and alternative approaches. Use when users need thorough analysis and structured planning before implementation, when breaking down complex features into actionable steps, or when they explicitly ask for a plan, roadmap, or strategy. Strictly planning-focused with no code modifications.

36121

project-planner

adrianpuiu

Comprehensive project planning and documentation generator for software projects. Creates structured requirements documents, system design documents, and task breakdown plans with implementation tracking. Use when starting a new project, defining specifications, creating technical designs, or breaking down complex systems into implementable tasks. Supports user story format, acceptance criteria, component design, API specifications, and hierarchical task decomposition with requirement traceability.

32115

system-design

lagz0ne

Use when designing, architecting, or planning a new system from requirements or ideas - transforms concepts into navigable design catalog using EventStorming methodology, Mermaid diagrams, and progressive elaboration through 5 phases (Requirements, Big Picture, Processes, Data/Flows, Integration)

19108

spec-kit-workflow

jmanhype

Guides specification-driven development workflow. Automatically invoked when discussing new features, specifications, technical planning, or implementation tasks. Ensures proper workflow phases (specify → clarify → plan → checklist → tasks → analyze → implement).

11111

sparc-methodology

ruvnet

SPARC (Specification, Pseudocode, Architecture, Refinement, Completion) comprehensive development methodology with multi-agent orchestration

8100

spec-workflow

TencentCloudBase

Standard software engineering workflow for requirement analysis, technical design, and task planning. Use this skill when developing new features, complex architecture designs, multi-module integrations, or projects involving database/UI design.

1091

Search skills

Search the agent skills registry