Whatif

Read the lead agent's full persona from their agent file and embody their perspective throughout.

Pre-flight

Read docs/pdlc/memory/STATE.md to capture current phase, sub-phase, feature, and last checkpoint.

If STATE.md does not exist, stop:

"PDLC not initialized. Run /pdlc init first."

Step 1 — Announce and set expectations

Read skills/formatting.md and output a Sub-phase Transition Header for "WHAT-IF ANALYSIS" followed by:

[Lead Agent Name] ([Role]): "Interesting question. Let me gather the team for a read-only analysis — no changes will be made, we're just exploring the implications. I'll present the findings when they're done."

If a phase is active (not Idle):

"Pausing [phase] / [sub-phase] at checkpoint [last checkpoint value].

If this session is interrupted:

• Run /pdlc whatif — PDLC will detect the pending analysis and offer to resume
• Run [resume command] — resumes the [phase] workflow from its last checkpoint"

If no phase is active:

"If this session is interrupted, re-run: /pdlc whatif [scenario text]"

Then output a Meeting Announcement Block (per skills/formatting.md):

• Called by: [Lead Agent Name] ([Role])
• Participants: all 9 agents
• Purpose: Read-only analysis of: "[scenario text]"
• Estimated time: ~2–4 minutes

Step 2 — What-If Analysis Meeting

Convene a party meeting following skills/build/party/orchestrator.md for spawn mode, agent roster, and the durable checkpoint protocol.

Write the pending party file per the orchestrator's durable checkpoint protocol with meetingType: "whatif-analysis".

2a — Individual agent assessment

Each agent evaluates the scenario hypothetically — what would happen if this were implemented? No files are read for modification purposes; agents assess from their knowledge of the current codebase and project state.

Each agent produces:

Agent: [name]
Impact: [None / Low / Medium / High]
Summary: [2-3 sentences on what would change]
Effort estimate: [Trivial / Small / Medium / Large / Redesign]
Risks: [specific risks, or "None identified"]

2b — Team discussion

After individual assessments, the team discusses:

1. Feasibility — Is this achievable with the current architecture? Or would it require a significant redesign?
2. Effort — Rough consensus on total effort (sum of all agents' estimates)
3. Risks — Cross-cutting risks that span multiple agents' domains
4. Trade-offs — What would we gain vs. what would we lose?
5. Recommendation — Team's overall take: pursue, defer, modify, or abandon

2c — Write MOM

Write the MOM to: docs/pdlc/mom/MOM_whatif_[scenario-slug]_[YYYY-MM-DD].md

Generate [scenario-slug] from the scenario text (first 3-5 words, kebab-case, max 30 chars).

# What-If Analysis: [Scenario title]

**Date:** [today YYYY-MM-DD]
**Scenario:** [full scenario text]
**Phase when asked:** [phase / sub-phase]
**Feature context:** [current feature, or — if none]
**Status:** Exploratory (read-only)

---

## Agent Assessments

### [Agent Name] ([Role])
- **Impact:** [None/Low/Medium/High]
- **Summary:** [2-3 sentences]
- **Effort estimate:** [Trivial/Small/Medium/Large/Redesign]
- **Risks:** [specific risks]

[repeat for each agent]

---

## Team Discussion

### Feasibility
[summary]

### Total Effort Estimate
[consensus estimate with breakdown]

### Cross-cutting Risks
[list]

### Trade-offs
- **Gains:** [what we'd get]
- **Costs:** [what we'd lose or spend]

### Team Recommendation
[Pursue / Defer / Modify / Abandon] — [rationale in 1-2 sentences]

---

## Next Steps
[What the user should consider — framed as options, not directives]

Update .pending-party.json: set "progress": "mom-written" and "momFile": "[path]".

Step 3 — Present findings

Synthesize the MOM into a concise summary for the user:

[Lead Agent Name] ([Role]): "Here's what the team found:

Scenario: [scenario text] Overall impact: [Low/Medium/High] Effort estimate: [consensus] Team recommendation: [Pursue/Defer/Modify/Abandon] — [one sentence rationale]

Key findings:

• [top 3-5 findings across all agents]

Risks:

• [top risks]

Full analysis: docs/pdlc/mom/MOM_whatif_[slug]_[date].md

What would you like to do?

• Explore further — dig deeper into a specific aspect
• Accept as decision — convert this into a formal decision (triggers the decision workflow)
• Discard — file the analysis and get back to where you left off"

Step 4 — Handle user's choice

Explore further

Ask: "What aspect would you like to explore deeper?"

Take the user's input and run another What-If Analysis Meeting (back to Step 2) with a refined scenario that incorporates the original findings plus the new angle. The new MOM is written as a separate file with a _v2, _v3 suffix.

Repeat until the user chooses Accept or Discard.

Accept as decision

The user wants to convert this exploration into a formal decision. Execute the decision workflow from skills/decision/SKILL.md with these modifications:

• Decision text: use the scenario text as the decision
• Source: User (explicit, via /pdlc whatif)
• Skip Step 2 (Decision Review Party): The What-If meeting already produced a comprehensive assessment. Do NOT convene another meeting.
• Use existing MOM: In the ADR entry, set MOM: to point to the What-If MOM file (or the latest version if multiple rounds were run). The impact summary in the ADR should be derived from the What-If MOM, not from a new assessment.
• Delete .pending-party.json before entering the decision skill (the whatif meeting is done).
• Run Step 1a of the decision skill (write .pending-decision.json) so the decision has crash recovery. Set "progress": "mom-written" since the assessment is already complete.
• Skip Step 1b (checkpoint display — already shown by whatif Step 1).
• Run Step 1c (classify — source: User (explicit, via /pdlc whatif)).
• Skip Step 2 (Decision Review Party — already done by whatif meeting).
• Start execution from Step 3: Present the What-If MOM findings as the impact assessment, ask the user to approve changes, then proceed through Steps 4 (record ADR), 5 (reconciliation), and 6 (summary and resume) normally.

This ensures:

• No duplicate meetings or MOM files
• Crash recovery via .pending-decision.json during Steps 3-6
• Step 6 reads STATE.md, informs the user it's resuming the paused workflow at Checkpoint A, and re-invokes the active phase's skill

Discard

[Lead Agent Name] ([Role]): "Analysis filed at docs/pdlc/mom/MOM_whatif_[slug]_[date].md for reference. Getting back to where we left off."

The MOM file is kept for future reference (it's read-only analysis, not a commitment).

Delete .pending-party.json.

Resume previous workflow:

Read STATE.md. If a phase was active:

"Resuming [phase] / [sub-phase] from checkpoint [last checkpoint value]."

Re-invoke the active phase's skill (/pdlc brainstorm, /pdlc build, or /pdlc ship). The skill reads STATE.md and resumes from the last checkpoint.

If no phase was active: stop.

Rules

• Read-only. The What-If skill never modifies project files (code, PRDs, design docs, ROADMAP.md, DECISIONS.md, STATE.md, Beads tasks). The only file it creates is the MOM.
• No decisions recorded unless the user explicitly chooses "Accept as decision."
• MOM files are permanent — even discarded analyses are kept for reference.
• When accepted as a decision, the What-If MOM is reused — never re-run the assessment.
• The durable checkpoint protocol applies: .pending-party.json is written before the meeting and deleted on completion.
• The lead agent mirrors the current phase lead, same as /pdlc decision.