frame-problem

Frame

Classify → route to right skill chain. Domain determines agent pattern, not just skill.

Framing: $ARGUMENTS

0. Auto-classify (skip if no $ARGUMENTS)

Read $ARGUMENTS. Attempt Cynefin domain classification using constraint language.

If confidence ≥80%: Propose classification — do NOT decide unilaterally.

🎯 Auto-classified: [Domain] (constraint: [type])
→ Verb: [probe|analyze|execute|act|decompose]
→ Suggested route: [skill chain]
Confirm? [Yes / Re-classify manually]

If confidence <80% or no $ARGUMENTS: Skip to Q1 (full qualification).

For Complicated domain with ≥80% confidence: also determine evolving vs degraded from context if possible. If clear → skip Q1.1 and route directly.

⚠️ AskUserQuestion Guard

CRITICAL: After EVERY AskUserQuestion call, check if answers are empty/blank. Known Claude Code bug: outside Plan Mode, AskUserQuestion silently returns empty answers without showing UI.

If answers are empty: DO NOT proceed with assumptions. Instead:

1. Output: "⚠️ Questions didn't display (known Claude Code bug outside Plan Mode)."
2. Present the options as a numbered text list and ask user to reply with their choice number.
3. WAIT for user reply before continuing.

1. Qualify

Question Refinement (before classifying): If $ARGUMENTS is vague or broad, generate 2-3 clarifying sub-questions that sharpen the problem statement. Present them inline before the AskUserQuestion call. Skip if $ARGUMENTS is already specific.

AskUserQuestion — both questions in one call:

Q1 "Constraints" (5 options, constraint-based):

• 🔒 Rigid — rules are fixed, process is known → Clear
• 📐 Governing — experts exist, best practices apply → Complicated
• 🌱 Enabling — experimentation needed, cause-effect unclear → Complex
• 🚨 Absent — no constraints, crisis or deliberate disruption → Chaotic
• ❓ Can't tell — none match or situation is mixed → Confused

Q1.1 "Complicated sub-question" (only if Q1 = Governing/Complicated):

• 📈 Evolving — system improving, capacity growing → /investigate
• 📉 Degraded — system failing, quality declining → /troubleshoot

(Skip Q1.1 if auto-classify already determined evolving vs degraded)

Q2 "Scale" (3 options, skip if Chaotic):

• 🪨 Boulder (multi-step, ambiguous, architectural)
• 🫧 Pebble (single file, obvious implementation)
• ❓ Not sure

2. Classify + Route

Map constraint type → domain → verb → skill chain:

Domain	Constraint	Verb	Scale	Route	OpenSpec?
Clear	Rigid	execute	Pebble	Just code it	No
Clear	Rigid	execute	Boulder	/openspec-develop directly	Yes
Complicated	Governing/Evolving	analyze	Any	/investigate → /openspec-plan	Boulder: yes
Complicated	Governing/Degraded	analyze	Any	/troubleshoot → stabilize → re-frame	No
Complex	Enabling/no hypothesis	probe	Any	/brainstorm → /probe → /openspec-plan	Yes
Complex	Enabling/has hypothesis	probe	Any	/probe → sense → /openspec-plan	Yes
Chaotic	Absent	act	—	/experiment → stabilize → /frame-problem	No
Confused	Unknown	decompose	Any	/frame-problem (re-classify after clarifying)	—

Present: 🎯 [Domain] → [Verb] → [skill chain] | OpenSpec: [yes/no] | Scale: [boulder/pebble]

3. Handoff

AskUserQuestion "Proceed?": Start chain / Re-frame / Skip framing.

On confirm → invoke first skill with $ARGUMENTS.