| name | review-plan |
| description | Reviews and edits all existing plan files based on user feedback. Accepts feedback from arguments (/review-plan change X to Y) or via AskUserQuestion. Same sequential flow as /plan (ui-ux -> tech-lead -> frontend+backend parallel) but EDITS existing plans instead of creating from scratch. Also updates progress.json if it exists. Triggers on: review plan, edit plan, update plan, change plan, revise plan, adjust plan. |
| user-invocable | true |
Review Plan Skill
Edit and update all existing plan files based on user feedback, following the same orchestration flow as /plan.
How It Differs from /plan
| Aspect | /plan | /review-plan |
|---|
| Purpose | Create plans from scratch | Edit existing plans |
| Input | PRD file | User feedback + existing plans |
| Agent action | Write new files | Read existing files, apply changes, rewrite |
| progress.json | Not created (use /complete-plan) | Updated/regenerated if it exists |
| Prerequisite | PRD must exist | Plan files must exist |
Flow
/review-plan [user feedback OR empty]
|
v
+----------------------------------------------------------+
| STEP 1: Get review feedback |
| - From arguments: /review-plan change auth to OAuth |
| - OR from AskUserQuestion if no arguments provided |
+----------------------------------------------------------+
|
v
+----------------------------------------------------------+
| STEP 2: Verify existing plan files |
| - docs/ui-ux-plan.md (REQUIRED) |
| - docs/dev-plan.md (REQUIRED) |
| - docs/frontend-plan.md (REQUIRED) |
| - docs/backend-plan.md (REQUIRED) |
| - docs/progress.json (optional โ update if exists) |
+----------------------------------------------------------+
|
v
+----------------------------------------------------------+
| STEP 3: ui-ux-specialist (SEQUENTIAL) |
| - Reads: docs/ui-ux-plan.md + user feedback |
| - EDITS and rewrites: docs/ui-ux-plan.md |
+----------------------------------------------------------+
|
v
+----------------------------------------------------------+
| STEP 4: tech-lead-specialist (SEQUENTIAL) |
| - Reads: updated docs/ui-ux-plan.md + user feedback |
| - EDITS and rewrites: docs/dev-plan.md |
+----------------------------------------------------------+
|
v
+----------------------------------------------------------+
| STEP 5: PARALLEL โ frontend + backend specialists |
| |
| frontend-specialist |
| - Reads: updated docs/dev-plan.md + user feedback |
| - EDITS and rewrites: docs/frontend-plan.md |
| |
| backend-specialist |
| - Reads: updated docs/dev-plan.md + user feedback |
| - EDITS and rewrites: docs/backend-plan.md |
+----------------------------------------------------------+
|
v
+----------------------------------------------------------+
| STEP 6: progress-creator (SEQUENTIAL, only if exists) |
| - Reads: updated dev-plan.md, frontend-plan.md, |
| backend-plan.md |
| - Rewrites: docs/progress.json |
+----------------------------------------------------------+
|
v
+----------------------------------------------------------+
| STEP 7: Report changes |
+----------------------------------------------------------+
Step 1: Get Review Feedback
Option A: Feedback from arguments
If the user invokes with arguments:
/review-plan change the auth flow to use OAuth instead of magic links
/review-plan add a dark mode toggle to the settings page
/review-plan remove the billing feature entirely
Extract the feedback directly from the arguments. Do NOT ask questions โ the user already told you what to change. Proceed directly to Step 2.
Option B: No arguments provided
If the user invokes with no arguments (/review-plan), you MUST use AskUserQuestion to gather feedback.
AskUserQuestion:
question: "What changes would you like to make to the plan?"
header: "Change Type"
options:
- label: "Add a feature"
description: "Add a new feature or capability to the plan"
- label: "Remove a feature"
description: "Remove an existing feature from the plan"
- label: "Change approach"
description: "Modify how a feature is implemented"
- label: "I'll describe it"
description: "Provide detailed feedback in my own words"
Then follow up with a specific question based on their choice:
If "Add a feature":
AskUserQuestion:
question: "What feature would you like to add? Describe it briefly."
header: "New Feature"
options:
- label: "New page/screen"
description: "Add a new page with its own route"
- label: "New component/widget"
description: "Add a new UI element to an existing page"
- label: "New backend capability"
description: "Add new tables, APIs, or data processing"
- label: "I'll describe it"
description: "Let me explain what I need"
If "Remove a feature":
AskUserQuestion:
question: "Which feature should be removed?"
header: "Remove"
options:
- label: "I'll specify it"
description: "Let me name the feature to remove"
If "Change approach":
AskUserQuestion:
question: "What would you like to change?"
header: "Change"
options:
- label: "UI/UX design"
description: "Change layout, colors, components, user flows"
- label: "Backend/database"
description: "Change schema, APIs, data model"
- label: "Tech stack"
description: "Change frameworks, libraries, tools"
- label: "I'll describe it"
description: "Let me explain the change"
Question Rules
- Maximum 3 questions to gather feedback
- If the user provided arguments, ask ZERO questions โ use their input directly
- Capture the full feedback as
{user_feedback} for all agent prompts
Step 2: Verify Existing Plan Files
Check that required plan files exist:
Glob: **/ui-ux-plan.md
Glob: **/dev-plan.md
Glob: **/frontend-plan.md
Glob: **/backend-plan.md
Glob: **/progress.json
If any REQUIRED file is missing:
Tell the user which files are missing and suggest running /plan first:
Missing: [file]. Run /plan first to create all plan files, then use /review-plan to edit them.
STOP. Do NOT proceed with partial plans.
Track progress.json existence:
has_progress_json = true/false
If progress.json exists, it will be regenerated in Step 6.
Step 3: Spawn ui-ux-specialist โ EDIT Mode (SEQUENTIAL)
Task:
subagent_type: ui-ux-specialist
description: "Edit UI/UX plan"
prompt: |
You are in REVIEW/EDIT mode โ NOT creating from scratch.
Read the EXISTING plan:
- docs/ui-ux-plan.md
User feedback to apply:
---
{user_feedback}
---
Instructions:
1. Read docs/ui-ux-plan.md completely
2. Identify which sections are affected by the user's feedback
3. Apply the requested changes to those sections
4. Keep all UNAFFECTED sections exactly as they are โ do not rewrite unchanged content
5. Add a "## Revision Log" section at the bottom with:
- Date of revision
- Summary of what changed
6. Write the updated file back to: docs/ui-ux-plan.md
IMPORTANT:
- This is an EDIT, not a rewrite. Preserve existing work.
- Only change what the user asked to change.
- If the feedback doesn't affect UI/UX (e.g., only backend changes), make no changes and return "No UI/UX changes needed."
Write the file using the Write tool. Return "Done: docs/ui-ux-plan.md [summary of changes]" when complete.
Wait for completion before proceeding.
Verify:
Glob: docs/ui-ux-plan.md
Step 4: Spawn tech-lead-specialist โ EDIT Mode (SEQUENTIAL)
Task:
subagent_type: tech-lead-specialist
description: "Edit dev plan"
prompt: |
You are in REVIEW/EDIT mode โ NOT creating from scratch.
Read the EXISTING plans:
- docs/dev-plan.md (your output โ edit this)
- docs/ui-ux-plan.md (may have been updated in previous step)
User feedback to apply:
---
{user_feedback}
---
Instructions:
1. Read docs/dev-plan.md completely
2. Read docs/ui-ux-plan.md to see if it was updated
3. Identify which phases and tasks are affected by:
a. The user's direct feedback
b. Any changes propagated from the updated ui-ux-plan.md
4. Apply changes:
- ADD new tasks if the feedback introduces new work
- REMOVE tasks if the feedback eliminates features
- MODIFY tasks if the feedback changes approach or scope
- UPDATE dependencies if task structure changed
5. Keep all UNAFFECTED phases and tasks exactly as they are
6. Add a "## Revision Log" section at the bottom with:
- Date of revision
- Summary of what changed (tasks added, removed, modified)
7. Write the updated file back to: docs/dev-plan.md
IMPORTANT:
- This is an EDIT, not a rewrite. Preserve existing work.
- Task IDs must remain consistent for any unchanged tasks.
- New tasks get the next available ID in their phase (e.g., if 3.1 and 3.2 exist, new task is 3.3).
- Removed tasks should be deleted entirely, not marked as removed.
- Do NOT create progress.json โ that is handled separately.
Write the file using the Write tool. Return "Done: docs/dev-plan.md [summary of changes]" when complete.
Wait for completion before proceeding.
Verify:
Glob: docs/dev-plan.md
Step 5: Spawn frontend + backend Specialists โ EDIT Mode (PARALLEL)
CRITICAL: Spawn BOTH agents in a SINGLE message with two Task tool calls.
[Single message with 2 Task tool calls]
Task 1: frontend-specialist
subagent_type: frontend-specialist
description: "Edit frontend plan"
prompt: |
You are in REVIEW/EDIT mode โ NOT creating from scratch.
Read the EXISTING plans:
- docs/frontend-plan.md (your output โ edit this)
- docs/dev-plan.md (may have been updated โ sync with it)
- docs/ui-ux-plan.md (may have been updated โ check design changes)
User feedback to apply:
---
{user_feedback}
---
Instructions:
1. Read docs/frontend-plan.md completely
2. Read docs/dev-plan.md to see what tasks changed
3. Read docs/ui-ux-plan.md to see if design specs changed
4. Identify which component specs are affected by:
a. The user's direct feedback
b. Any tasks added/removed/modified in dev-plan.md
c. Any design changes in ui-ux-plan.md
5. Apply changes:
- ADD new component specs for new tasks
- REMOVE specs for deleted tasks
- MODIFY specs where approach or design changed
- UPDATE the Component Index table
6. Keep all UNAFFECTED specs exactly as they are
7. Add a "## Revision Log" section at the bottom with:
- Date of revision
- Summary of what changed
8. Write the updated file back to: docs/frontend-plan.md
IMPORTANT:
- This is an EDIT, not a rewrite. Preserve existing work.
- Maintain Bulletproof React structure in all new/modified specs.
- If the feedback doesn't affect frontend, make no changes and return "No frontend changes needed."
Write the file using the Write tool. Return "Done: docs/frontend-plan.md [summary of changes]" when complete.
Task 2: backend-specialist
subagent_type: backend-specialist
description: "Edit backend plan"
prompt: |
You are in REVIEW/EDIT mode โ NOT creating from scratch.
Read the EXISTING plans:
- docs/backend-plan.md (your output โ edit this)
- docs/dev-plan.md (may have been updated โ sync with it)
User feedback to apply:
---
{user_feedback}
---
Instructions:
1. Read docs/backend-plan.md completely
2. Read docs/dev-plan.md to see what tasks changed
3. Identify which schema, migration, or RLS specs are affected by:
a. The user's direct feedback
b. Any tasks added/removed/modified in dev-plan.md
4. Apply changes:
- ADD new table/migration specs for new backend tasks
- REMOVE specs for deleted tasks
- MODIFY schemas, RLS policies, or queries where approach changed
- UPDATE migration order if table structure changed
- UPDATE schema diagram
5. Keep all UNAFFECTED specs exactly as they are
6. Add a "## Revision Log" section at the bottom with:
- Date of revision
- Summary of what changed
7. Write the updated file back to: docs/backend-plan.md
IMPORTANT:
- This is an EDIT, not a rewrite. Preserve existing work.
- Maintain Supabase best practices in all new/modified specs.
- If the feedback doesn't affect backend, make no changes and return "No backend changes needed."
Write the file using the Write tool. Return "Done: docs/backend-plan.md [summary of changes]" when complete.
Wait for BOTH agents to complete before proceeding.
Step 6: Update progress.json (CONDITIONAL)
Only run this step if progress.json existed in Step 2.
If has_progress_json == true:
Task:
subagent_type: progress-creator
description: "Update progress.json"
prompt: |
Read all UPDATED plan files and REGENERATE progress.json.
REQUIRED โ read this first:
- docs/dev-plan.md (has been updated)
OPTIONAL โ read these for enrichment (if they exist):
- docs/frontend-plan.md (may have been updated)
- docs/backend-plan.md (may have been updated)
Your job:
1. Parse every phase and task from the UPDATED dev-plan.md
2. Enrich tasks with file paths and details from specialist plans
3. Classify each task as "low" or "high" complexity with a specific reason
4. Assign agent: "small-coder" for low, "high-coder" for high
5. Map dependencies between tasks
6. Calculate summary counts
7. All task statuses reset to "pending" (fresh regeneration)
Write the output to: docs/progress.json
Use the Write tool to create the file. Return "Done: docs/progress.json" when complete.
If has_progress_json == false:
Skip this step. Tell the user:
progress.json was not found โ skipping regeneration. Run /complete-plan to generate it when ready.
Step 7: Report Changes
After all agents have completed, provide a summary:
## Plan Review Complete
### Changes Applied
**User request:** {user_feedback}
### Files Updated
| # | File | Agent | Status |
|---|------|-------|--------|
| 1 | docs/ui-ux-plan.md | ui-ux-specialist | [Updated / No changes needed] |
| 2 | docs/dev-plan.md | tech-lead-specialist | [Updated / No changes needed] |
| 3 | docs/frontend-plan.md | frontend-specialist | [Updated / No changes needed] |
| 4 | docs/backend-plan.md | backend-specialist | [Updated / No changes needed] |
| 5 | docs/progress.json | progress-creator | [Regenerated / Skipped] |
### Execution Order
Step 1: Gathered user feedback
Step 2: Verified plan files exist
Step 3: ui-ux-specialist -> edited docs/ui-ux-plan.md
Step 4: tech-lead-specialist -> edited docs/dev-plan.md
Step 5: frontend-specialist -> edited docs/frontend-plan.md } PARALLEL
backend-specialist -> edited docs/backend-plan.md }
Step 6: progress-creator -> regenerated docs/progress.json (if existed)
### What Changed
Each updated file has a **Revision Log** at the bottom documenting:
- Date of revision
- Summary of specific changes made
### Next Steps
1. Review the Revision Log in each updated file
2. If progress.json was regenerated, verify task classifications
3. Run `/developer` to continue executing tasks
Error Handling
Missing plan files
STOP. List which files are missing.
Tell user: "Run /plan first to create plan files."
Agent reports "No changes needed"
This is valid โ the user's feedback may not affect all domains.
Report which files were unchanged and why.
Agent fails
1. Report which agent failed and the error
2. Other agents' changes are still valid
3. User can re-run /review-plan for the same feedback to retry
Execution Rules
DO:
- ALWAYS check for arguments first โ skip AskUserQuestion if feedback is in arguments
- ALWAYS use AskUserQuestion if NO arguments are provided
- ALWAYS verify all 4 plan files exist before spawning any agents
- ALWAYS pass
{user_feedback} to EVERY agent โ they need context to know what changed
- ALWAYS run ui-ux first, then tech-lead, then frontend+backend in parallel
- ALWAYS update progress.json if it existed before the review
- ALWAYS tell agents this is EDIT mode, not creation from scratch
- ALWAYS ask agents to add a Revision Log
DO NOT:
- NEVER rewrite plans from scratch โ this is an edit operation
- NEVER skip the sequential order (ui-ux -> tech-lead -> frontend+backend)
- NEVER spawn frontend/backend before tech-lead completes
- NEVER ask more than 3 questions to gather feedback
- NEVER ask questions if the user provided arguments
- NEVER pass file content in prompts โ use file paths only
- NEVER modify plan files yourself โ let the agents do it
Checklist
