ck:plan

ck:plan — Structured Planning Pipeline

---

Step 0 — Scope Challenge

Before spawning any agents, detect mode and challenge scope:

# Scope Challenge:
#   Exists?     → [does this feature already exist in the codebase?]
#   Minimum?    → [smallest impl that satisfies requirements]
#   Complexity? → [Fast | Hard] — reasons: multi-file? unfamiliar? security?
#
# Mode: [Fast | Hard]
# Test:  [default | --tdd]

Mode auto-detection (override with explicit flag):

• Fast — single-file change, familiar pattern, ≤ 2 components
• Hard — multi-file, unfamiliar domain, security-sensitive, or ≥ 3 phases

If scope is too large: suggest splitting and wait for user confirmation.

If Hard and novel/ambiguous with no brainstorm report: "No brainstorm found. Run /ck:brainstorm first? [Y/n]" — if Yes, stop; if No, proceed.

If a spec file path is provided or plans/{slug}/spec.md exists adjacent to any plan: run a Spec Quality Check inline:

# Spec Quality Check:
#   [NEEDS CLARIFICATION] remaining? → CRITICAL — resolve before continuing
#   Success criteria measurable?     → HIGH if vague adjectives (fast, scalable, reliable)
#   User stories P1/P2/P3?           → HIGH if missing
#   Acceptance criteria testable?    → MEDIUM if vague ("works correctly")
#
# Verdict: [PASS | WARN (list) | BLOCK (list)]

• BLOCK: surface findings, resolve before proceeding
• WARN: list findings, user acknowledges — then proceed
• PASS: continue normally

---

Step 1 — Research (Hard only)

Spawn 2 researcher agents in parallel:

• Instance A — role: Primary — recommended approach and best practices
• Instance B — role: Alternative — alternative approach and tradeoffs

// Researcher A (Primary): [approach] → [verdict]
// Researcher B (Alternative): [approach] → [verdict]

---

Step 2 — Plan Creation

Spawn the planner agent with: feature description + mode + research reports + test flag + spec file path (if any).