Installs into .claude/skills of the current project.
Are you the author of 3 Plan?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/tstapler-3-plan)
---
description: "Phase 3 — Task breakdown, materials list, and budget. Outputs: home_plans/<project>/plan.md"
user-invocable: true
---
# home:3-plan
Dispatch a planning subagent to produce the full project plan — task sequence, materials list, budget, and timeline. Then run an adversarial reviewer to catch gaps before work begins.
## Instructions
1. **Follow [SETUP.md](SETUP.md)** — identify PROJECT_NAME, and mark this stage's task `in_progress` if a task tool is available (Check 3).
2. **Entry gate — read all inputs** (coordinator reads these to pass to subagent):
- `home_plans/<PROJECT_NAME>/scope.md` — halt if missing, run `/home:1-scope` first
- `home_plans/<PROJECT_NAME>/research/methods.md` — warn if missing, continue with scope only
- `home_plans/<PROJECT_NAME>/research/materials.md`
- `home_plans/<PROJECT_NAME>/research/logistics.md`
- `home_plans/<PROJECT_NAME>/research/risks.md`
3. **Dispatch a planning subagent using the `Agent` tool.**
The subagent prompt must include:
- Full text of `scope.md`
- Full text of all research files (if present)
- These exact instructions:
> You are a home project planning subagent. Produce a complete, actionable project plan grounded in the research provided — do not invent facts the research doesn't support.
>
> **Step 1:** Review the scope and research. Identify whether any permits are required — if so, note them prominently at the top of the plan.
>
> **Step 2:** Write `home_plans/<PROJECT_NAME>/plan.md` following the template below. Use concrete product names and quantities — no vague references. Tasks should be sized for a single work session (30 min–3 hrs each). Where a research file linked a product to a wiki page (`[[Product Name]]`), reuse that same wiki-link in the Materials List rather than plain text — this keeps the plan connected to the wiki's product/retailer graph. Carry forward source citations from the research files: every non-obvious budget, code, or safety claim in the plan should be traceable to a source in the research, listed in the plan's trailing `## Sources` section.
>
> **Step 3:** Return a summary: phase count, task count, total budget estimate, whether permits are required.
Plan template:
```markdown
# Project Plan: <PROJECT_NAME>
**Date**: <YYYY-MM-DD>
**Project**: <one-line description>
**Status**: Ready to execute
**Permits Required**: Yes/No — <type and estimated cost if yes>
**Estimated Total Cost**: $<low>–$<high>
**Estimated Duration**: <X days / X weekends>
---
## ⚠️ Safety Precautions
<Non-negotiable safety steps — utilities to shut off, PPE required, hazards to avoid>
---
## Task Sequence
### Phase 1: Prep & Procurement
#### Task 1.1: <name> (~<duration>)
- <Step-by-step actions>
- **Done when**: <specific observable completion criterion>
#### Task 1.2: <name> (~<duration>)
...
### Phase 2: <Core Work Phase Name>
#### Task 2.1: <name> (~<duration>)
...
### Phase 3: Finishing & Cleanup
#### Task 3.1: <name> (~<duration>)
...
---
## Materials List
| Item | Spec / Model | Qty | Unit Cost | Total | Source |
|------|-------------|-----|-----------|-------|--------|
| <item> | <specific product or spec> | <N> | $<X> | $<X> | Lowe's / Amazon / etc. |
**Materials subtotal**: $<X>
---
## Tools Required
| Tool | Own / Rent / Buy |
|------|-----------------|
| <tool> | Own |
| <tool> | Rent from Home Depot (~$<X>/day) |
---
## Budget Breakdown
| Category | Estimate |
|----------|---------|
| Materials | $<X> |
| Tool rental | $<X> |
| Permits | $<X> |
| Contractor labor (if any) | $<X> |
| Contingency (15%) | $<X> |
| **Total** | **$<X>–$<X>** |
---
## Timeline
| Day / Weekend | Work |
|--------------|------|
| Day 1 | <tasks> |
| Day 2 | <tasks> |
---
## Contractor Requirements
<If any work requires a licensed contractor: what license, what scope, and what to verify before hiring>
<"None required" if fully DIY>
---
## Decisions & Open Questions
<Any choices that need to be made before or during execution>
---
## Sources
<URLs carried forward from the research files, backing the budget/code/safety claims above>
```
4. **Wait for the subagent to complete.** Do not continue until plan.md has been written.
5. **Dispatch an adversarial reviewer subagent using the `Agent` tool.**
The subagent prompt must include:
- Full text of `plan.md`
- Full text of `scope.md`
- These exact instructions:
> You are an adversarial reviewer for a home project plan. Find gaps before any work begins.
>
> Review for:
> 1. **Missing safety steps** — Are there hazards (electrical, plumbing, structural, chemical) without explicit precautions? Should utilities be shut off? Is PPE specified?
> 2. **Permit gaps** — Does the scope require a permit that isn't noted? (Electrical work, structural changes, plumbing changes, HVAC modifications often require permits.)
> 3. **Budget underestimates** — Are costs realistic? Are common hidden costs (demo surprises, disposal fees, primer coats, drying time, re-work) accounted for? Is the contingency sufficient?
> 4. **Sequencing errors** — Are any tasks ordered incorrectly? (e.g., painting before priming, installing before measuring, ordering materials without lead time buffer)
> 5. **Missing materials** — Are consumables (sandpaper, tape, drop cloths, fasteners, caulk, primer) accounted for?
> 6. **Scope creep** — Are any tasks broader than what the scope asked for?
> 7. **Pet and occupant safety** — Are there steps requiring the space to be vacated? Toxic fumes? Sharp debris?
> 8. **Missing citations** — Are there budget, code, or safety claims stated as fact with no source in the plan's `## Sources` section or the underlying research?
>
> For each concern, classify as:
> - **BLOCKER** — Must be resolved before work starts (safety hazard, required permit, impossible task sequence)
> - **CONCERN** — Should be addressed; will likely cause problems if ignored
> - **MINOR** — Low impact; note it but don't block
>
> Write findings to `home_plans/<PROJECT_NAME>/adversarial-review.md`:
>
> ```markdown
> # Adversarial Review: <PROJECT_NAME>
>
> **Date**: <YYYY-MM-DD>
> **Verdict**: BLOCKED / CONCERNS / CLEAN
>
> ## Blockers
> - [ ] <issue> — <recommendation>
>
> ## Concerns
> - [ ] <issue> — <recommendation>
>
> ## Minors
> - <issue>
> ```
>
> Return a one-line summary: verdict + count of blockers/concerns/minors.
6. **Wait for the adversarial reviewer to complete.** Read the summary.
- **BLOCKED** → read adversarial-review.md, patch plan.md to resolve each BLOCKER, then re-run the adversarial reviewer on the updated plan (repeat until CONCERNS or CLEAN).
- **CONCERNS or CLEAN** → proceed.
7. **Exit gate**: confirm `plan.md` and `adversarial-review.md` both exist and are non-trivial, and that the review verdict is CONCERNS or CLEAN (never advance past a BLOCKED verdict). Mark this stage's task `completed` if a task tool is available, then output the coordinator summary:
```
✅ Phase 3 complete — plan.md written to home_plans/<PROJECT_NAME>/
Phases: <N> | Tasks: <N>
Budget estimate: $<X>–$<X>
Permits required: Yes/No
Adversarial review: <BLOCKED|CONCERNS|CLEAN> — <N> blockers, <N> concerns, <N> minors
Next step: /home:4-prep
```