plan-new

Create New Plan

You are creating a plan for: $ARGUMENTS

Follow the Universal Planning Framework. Execute ALL stages in order. No stage may be skipped or replaced by inline text.

STRICT READ-ONLY RULE

During /plan-new execution, ONLY the plan file may be edited. No other files may be created, edited, or deleted. No Bash commands that make changes. Only: Read, Grep, Glob, Explore/Plan Agents, AskUserQuestion, Skill (for interview/refine/review), TodoWrite (for phase/task tracking). Violation = abort planning.

This replaces EnterPlanMode/ExitPlanMode to avoid repeated UI previews that clutter the screen.

---

Step 0: Planning Mode Selection

Before starting, ask the user how they want to work:

AskUserQuestion:
  "How would you like to work through the plan?"
  Options:
    1. "Manual - I'll go through each phase step by step"
       (User controls each phase, chooses interactive or self-interview)
    2. "Autonomous - you handle everything independently"
       (All phases autonomous, interview in self-mode)
    3. "Autonomous with interactive interview"
       (All phases autonomous, but I answer the interview myself)

Store the selected mode and follow it through all stages.

---

Stage 0.pre: DSV (Quick Reasoning Check)

MANDATORY - Before any planning, run DSV on the task itself:

1. Decompose: What are the 2-3 key claims in this request?
2. Suspend: What alternative interpretation haven't I considered?
3. Validate: Which claim am I least sure about? Validate that first.

If Suspend reveals a plausible alternative: raise it before proceeding. Document DSV results in the plan file under ## DSV Pre-Check.

---

Stage 0: Discovery (Before Planning)

Determine which of the 12 checks to run. Use the intelligence rule: skip checks you can articulate a reason for skipping.

Priority tiers:

• Always: 0.1 Existing Work, 0.7 Feasibility, 0.9 Better Alternatives (AHA Effect)
• Always (coding domains): + 0.3 Official Docs
• Usually: 0.2 Factual Verification, 0.3 Official Docs, 0.8 ROI
• Context-dependent: 0.4 Updates Scan, 0.5 Best Practices, 0.6 Deep Research, 0.10 Competitive, 0.11 Constraint Discovery, 0.12 People Risk