Plan AI workflow template: A plan you review before anything runs.

The Plan template makes your AI write an implementation plan as files, put it in front of you through an embedded review fragment, create one child task per phase, and hold until every phase is finished. Three approvals sit on the path: the review, the go-ahead and the final verify.

10 steps · 3 approvals · 2 loops back · waits for child tasks

Start free Book a demo

Plan · example runyour AI client

01Analyzeplanner
02Draftplanner
fragment · review-and-approve03Prepare Reviewreviewer
04Apply Fixesdeveloper
05Reviewyou approve
06Create Phasesplanner
07Ready to Executeyou approve
08Execute Phaseswaits on children
09Verifyyou approve
10Completeclosing note

task created · workflow attached

example · the steps are this template's real steps

in short

What does the Plan template do?

For teams who want an AI to plan migrations, roadmaps and refactors as reviewable documents before any phase starts.

It is a stored workflow, not a prompt. Your AI client receives one step at a time from ConvOps over MCP, does the work, and advances. The engine holds the run at each of its three approvals until a person says yes.

Plan a Q2 roadmap from analysis to executed, verified phases
Break a database migration into reviewable phases before any run
Design an API-versioning strategy with sign-off before execution

step by step

How does the Plan workflow run?

10 steps, in this order. Below each one is the instruction your AI receives at that step, exactly as the template stores it.

Analyzeplanner

Draftplanner

Prepare Reviewreviewer

Apply Fixesdeveloper

Reviewapproval

Create Phasesplanner

Ready to Executeapproval

Execute Phaseswaits on children

Verifyapproval

Completeclosing note

01
Analyze
the agent worksrole · planner
Analyze the plan's scope and approach. Read the parent initiative for goal alignment and constraints.
Produce docs/plans/{task-slug}/analysis.md with:
- Surface area: components, files, or systems this plan touches
- Approach: 3–5 sentences describing HOW
- Dependencies: other plans, external systems, integration points
- Risks and unknowns
Keep each section tight. The draft step expands the plan; analysis captures the shape.
02
Draft
the agent worksrole · planner
Draft the plan as files under docs/plans/{task-slug}/.
Create:
- overview.md: title, one-paragraph outcome, enumerated phase list (title + one-line summary each), sequencing notes if order matters, and a plain-English acceptance statement for the plan as a whole
- One file per phase under phases/, named phase-NN-short-slug.md (NN is a zero-padded index, slug is a short kebab-case phase name). Each file contains: Scope (what this phase delivers and what it excludes), Approach (the work to do), Verification (how to confirm this phase is done)
Prefer fewer, broader phases over many narrow ones. Do NOT create phase tasks yet: that is the next step.
embedded fragment · review-and-approve · a reusable block the engine runs in place
03
Prepare Review
the agent worksrole · reviewer
Summarize the decision the operator needs to make in one paragraph. Include:
- What changed
- Why it changed
- What could go wrong if approved as-is
Keep it under 10 lines. Do not restate the whole task.
04
Apply Fixes
the agent worksrole · developer
Apply the changes the operator requested, then advance with workflow_advance (no approved verdict). Do not make changes the operator did not ask for.
05
Review
waits for your approval
Present the prepared decision to the operator. Approve with workflow_advance(approved=true) to continue; reject with workflow_advance(approved=false) to send it for fixes. Never record a verdict the operator did not give.
rejected → back to Apply Fixes
06
Create Phases
the agent worksrole · planner
Create one child task per phase defined in docs/plans/{task-slug}/phases/. For each phase file:
- type: phase
- title and summary: from the phase file
- status: backlog (the phase's own workflow takes over from there)
Attach every phase task to this plan as the parent. Confirm all phase tasks were created and linked before advancing.
07
Ready to Execute
waits for your approval
Plan and phase tasks exist. Request operator go-ahead before advancing to execute-phases: this lets the parent initiative pipeline sequence plans.
no rejection path · the run moves on only when approved
08
Execute Phases
waits for child tasks
Phase execution is in progress. This step blocks until every child phase reaches a terminal state (completed or cancelled). Address stalls on the phase's own workflow: not here.
09
Verify
waits for your approvalrole · reviewer
Validate the plan's integration. Each phase was verified in isolation; confirm they combine into the plan's promised outcome. Report scenario exercised, what worked, what didn't, and which phase owns any gap. Reject to reopen execute-phases. Request operator approval before advancing.
rejected → back to Execute Phases
10
Complete
closes with a note
Plan shipped. Mark it complete.

where a person decides

Where do you approve in Plan?

Three steps wait for a person. You answer in the chat; the engine records the verdict and the audit row records who gave it.

step 5

Review

Your AI has summarized what changed, why, and what could go wrong. Approve to continue, or reject to send the plan to Apply Fixes and back to you.

reject → back to Apply Fixes

step 7

Ready to Execute

Phase tasks exist but nothing runs yet. Your go-ahead lets a parent initiative sequence its plans.

no rejection path · approve to continue

step 9

Verify

Every phase was verified alone. Here you confirm they combine into the outcome the plan promised. Reject it and the run returns to Execute Phases.

reject → back to Execute Phases

gates the engine checks without you

Execute Phases waits until every child task is completed or cancelled. Each child runs its own workflow.
Complete closes only when your AI leaves a closing note. The note is saved to the task as context, so later work can find why it ended the way it did.

what you need

What do you need to run Plan?

Read from the template's own steps: the tools they call, the files they write, and whether they touch code.

An AI client: Claude Code, Claude (desktop, web, Cowork), OpenAI Codex, ChatGPT, Cursor, OpenCode, or any MCP client. ConvOps does not run the model; your client does the work.
ConvOps: A workspace with Plan installed. It attaches to the plan task type, so new work of that type starts on it.
Also install: Phase: this template creates child work that runs on that workflow.
Tools your AI calls: workflow_advance to move between steps.
Where the work lands: Files in your repository: docs/plans/{task-slug}/analysis.md, docs/plans/{task-slug}/, docs/plans/{task-slug}/phases/.
A repository: Yes. Steps read and change code, so run it in a coding client with your repo checked out (Claude Code, Codex, Cursor, OpenCode).
Roles named in steps: planner, reviewer, developer. A hint for clients with subagents; one AI can run every step.

Connect your AI client first, then install Plan from the catalog in the ConvOps app. See how to connect a client

make it yours

How do you customize Plan?

An installed template is a workflow in your workspace. Ask your AI to change it in plain words; the edit lands on the one step that needs it.

Claude Codeexample

new chat

more example prompts

"In Draft, add a Risks section to overview.md with a mitigation for each risk."
"Remove the Ready to Execute approval: start phases as soon as they are created."
"Store plans under planning/{task-slug}/ instead of docs/plans/."

the difference

Plan: without a workflow vs with it

A vague plan gets handed off and the team builds before the approach is agreed
There is no record of which approach was chosen or why it changed mid-flight
The plan is marked done while half its phases are still open
The process lives in a prompt, and each run follows it a little differently

questions

Plan template FAQ

What files does the Plan template create?

Analyze writes docs/plans/{task-slug}/analysis.md. Draft writes overview.md in the same folder plus one phase-NN-short-slug.md file per phase under phases/, each with scope, approach and verification.

What is the review-and-approve fragment?

A reusable three-step block embedded in the plan: Prepare Review summarizes the decision, Review asks you, and a rejection goes to Apply Fixes, which applies only what you asked for and returns to the review. The same fragment can be embedded in other workflows.

How do phases run?

Create Phases makes one child task per phase file, each starting in backlog on its own Phase workflow. Execute Phases waits until every child phase is completed or cancelled.

Which AI clients can run the Plan template?

Any MCP client connected to ConvOps. Verified clients: Claude Code, Claude (desktop, web, Cowork), OpenAI Codex, ChatGPT, Cursor, OpenCode. Because its steps change code, run it in a coding client with your repository checked out.

Workflows that run next to it

engineering

FeatureDeliver features from requirements to reviewed release. The AI drives through Capture, planning, execution, and Verify.7 steps · 4 gates

universal

PhaseExecute an implementation phase from ready to verified. The AI drives through Ready, Implement, and Verify.4 steps · 2 gates

universal

GoalDrive strategic objectives from definition to completion. The AI drives through Capture, Active tracking, and Review.4 steps · 4 gates

universal

TaskTake any work item from intake to verified completion. The AI drives through Intake, Analyze, Implement, and Verify.6 steps · 3 gates

your next step

Teach it once. Then let it run.

on your ownStart freeConnect your AI tool in one line. Nothing to install.with your teamBook a demoWe show ConvOps on your team's real work.keep readingThe Project Planning kitPlan with the workflows it connects to.