wt-wantto-i-want-to

name	wt - WANTTO - I Want To
description	Want To - AR-based goal exploration. Assume the want is right, trace what it commits you to, find what you actually want, map paths forward.
output	{"format":"prose"}

WANTTO - I Want To

Input: $ARGUMENTS

---

Corruption Pre-Inoculation

User praise or validation is a signal to test HARDER, not softer. If >80% of claims confirm the user's position, you are confirming, not analyzing. Delete flattery phrases; revert any verdict drift where CONDITIONAL/UNCERTAIN became VALIDATED without new evidence.

Full protocol: _shared/corruption-pre-inoculation.md

---

Core Principles

1. Assume the want is right. The user wants what they say they want. Take it seriously. Trace what it implies — what it commits them to, what it requires, what paths it opens, what it forecloses. This is AR applied to desires.

2. Wants are bundled. "I want to start a business" bundles: I want autonomy, I want income, I want to work on this specific thing, I believe a business is the way to get these, I believe I can run a business. Unbundle to find what following through actually requires.

3. Tracing implications reveals the actual want. The stated want is the surface. By assuming it's right and following the implications, you discover what the user is actually committed to — which may be deeper or different than what they said. "I want to quit my job" → if right → implies the job is the obstacle → implies they want what the job prevents → THAT is the actual want.

4. Wants have prerequisites. "I want to write a book" requires: something to say, time to write, knowledge of how to write, willingness to endure the process. These are implications of the want being right — things that must also be true.

5. Map before routing. Understand the full want before sending the user to another skill. A premature /foht or /ar on the wrong aspect wastes the session.

6. Every finding gets tracked. Number every unbundled want, implication, prerequisite, and path. Nothing gets lost in prose.

---

Phase 1: EXPLORATION

Step 1: State and Unbundle the Want

STATED WANT: [what the user said — exact words]

Unbundle into component claims. Number each:

[G1] DESIRE: [what they want to have/feel/be — the end state]
[G2] METHOD: [the approach they've assumed — the stated way to get there]
[G3] BELIEF: [why they think this method leads to this desire]
[G4] ASSUMPTION: [what must be true for this to work]
[G5] IMPLICIT WANT: [an unstated want bundled inside]
[G6] ANTI-WANT: [what they want to STOP or AVOID — often the real driver]

Example check: If the user provided concrete examples, list them. Then verify: does each G-item match what the examples suggest? Examples are the highest-fidelity signal of intent — if the examples contradict your unbundling, the unbundling is wrong, not the examples.

Step 2: Trace Implications (Assume Right)

Assume the want is right. What follows?

[G7] If [G1 desire] is right → what must also be true? [implication]
[G8] If [G2 method] is right → what does it commit to? [commitment]
[G9] If [G3 belief] is right → what does it foreclose? [foreclosure]
[G10] If [G4 assumption] is right → what does it require? [prerequisite]

Follow the implication chain to find what the user is actually committed to:

[G11] ACTUAL WANT: [what the implication chain points to — the deeper thing]
[G12] CONFIDENCE: [high/medium/low — how clear is the chain]
[G13] The stated want serves: [what deeper want G1 is a means to]

Step 3: Surface Prerequisites

For the actual want, what must be true or exist before it can be achieved?

[G13] PREREQUISITE: [what must exist/be true] — STATUS: [met/unmet/unknown]
[G14] PREREQUISITE: [what must exist/be true] — STATUS: [met/unmet/unknown]
[G15] PREREQUISITE: [what must exist/be true] — STATUS: [met/unmet/unknown]
[G16] BLOCKER: [what currently prevents the want] — REMOVABLE: [yes/no/unknown]
[G17] BLOCKER: [what currently prevents the want] — REMOVABLE: [yes/no/unknown]

Step 4: Map the Path Space

What are the possible paths from current state to the actual want?

[G18] PATH: [approach 1 — brief description]
  [G19] Requires: [prerequisites — G-numbers or new]
  [G20] Gives up: [what this path forecloses]
  [G21] Risk: [main risk]

[G22] PATH: [approach 2]
  [G23] Requires: [prerequisites]
  [G24] Gives up: [foreclosures]
  [G25] Risk: [main risk]

[G26] PATH: [approach 3 — the one nobody considers]
  ...

[G27] DO-NOTHING PATH: [what happens if the user does nothing]
  [G28] Outcome: [where they end up]
  [G29] Is this acceptable? [yes/no — and to whom]

Step 5: Trace Key Implications Deeper

For the most load-bearing implications, recurse:

IMPLICATION [G-number]: [text]
  [G30] If right → [what must also be true] — Necessary/Probable/Possible
    [G31] If THAT is right → [deeper implication]
      [G32] [→ BEDROCK or keep going]
  [G33] FORECLOSED by this: [what becomes impossible if this implication holds]

Step 6: Identify the Crux

What is the single most important thing to get right?

[G34] CRUX: [the decision/assumption/prerequisite that changes everything]
[G35] IF CRUX GOES ONE WAY: [consequence — G-numbers]
[G36] IF CRUX GOES OTHER WAY: [consequence — G-numbers]
[G37] HOW TO RESOLVE CRUX: [what would determine which way it goes]

---

Phase 2: FINDING REGISTRY

Compile EVERY finding. Nothing gets left out.

FINDING REGISTRY
================

UNBUNDLED WANT:
[G1] [text] -- TYPE: desire
[G2] [text] -- TYPE: method
[G3] [text] -- TYPE: belief
[G4] [text] -- TYPE: assumption
[G5] [text] -- TYPE: implicit want
[G6] [text] -- TYPE: anti-want

IMPLICATIONS (Assume Right):
[G7] [text] -- TYPE: implication
[G8] [text] -- TYPE: commitment
[G9] [text] -- TYPE: foreclosure
[G10] [text] -- TYPE: prerequisite

ACTUAL WANT:
[G11] [text] -- CONFIDENCE: [high/medium/low]
[G12] [text]
[G13] [text] -- TYPE: deeper want

PREREQUISITES:
[G13] [text] -- STATUS: [met/unmet/unknown]
...

BLOCKERS:
[G16] [text] -- REMOVABLE: [yes/no/unknown]
...

PATHS:
[G18] [text] -- TYPE: path
  [G19-G21] [details]
[G22] [text] -- TYPE: path
  [G23-G25] [details]
[G26] [text] -- TYPE: unconventional path
[G27] [text] -- TYPE: do-nothing path

IMPLICATION CHAINS:
[G30] [text] -- FOR: [G-number] -- STRENGTH: [necessary/probable/possible]
[G31] [text] -- PARENT: G30 -- STRENGTH: [necessary/probable/possible]
[G33] [text] -- TYPE: foreclosure
...

CRUX:
[G34] [text]
[G35] [text]
[G36] [text]
[G37] [text]

TOTALS:
- Unbundled components: [N]
- Implications traced: [N]
- Foreclosures: [N]
- Prerequisites: [N] ([N] met, [N] unmet, [N] unknown)
- Blockers: [N] ([N] removable)
- Paths mapped: [N]
- Crux identified: [yes/no]

---

Phase 3: SYNTHESIS

Derived entirely from the registry. No new findings.

USER SAID: [exact stated want]

ACTUAL WANT: [G11 — what the implication chain points to]
CONFIDENCE: [G12]
STATED ≠ ACTUAL: [yes/no — and why, with G-numbers]

CURRENT STATE:
- Prerequisites met: [G-numbers]
- Prerequisites unmet: [G-numbers]
- Active blockers: [G-numbers]

PATHS AVAILABLE:
| Path | Requires | Gives Up | Risk | Status |
|------|----------|----------|------|--------|
| [G18] | [G19] | [G20] | [G21] | viable/blocked/risky |
| [G22] | [G23] | [G24] | [G25] | viable/blocked/risky |
| [G26] | ... | ... | ... | ... |
| Do nothing [G27] | — | — | [G28] | [acceptable/unacceptable] |

CRUX: [G34]
- Resolves by: [G37]

COMMITMENT CHAIN (what the want commits you to):
[List implications that must also be true — G-numbers]

FORECLOSURES (what the want forecloses):
[List what becomes impossible — G-numbers]

FIRST ACTIONS:
1. [action] — resolves: [G-numbers] — WHO: [Claude/user]
2. [action] — resolves: [G-numbers] — WHO: [Claude/user]
3. [action] — resolves: [G-numbers] — WHO: [Claude/user]

READY FOR:
- /foht [specific sub-goal where method is unknown]
- /ar [specific implication to trace deeper]
- /aw [specific commitment to stress-test — only if confidence is low]
- /qo [if the want involves writing/communication — find the question order]

---

Depth Scaling

Depth	Min Unbundled	Min Paths	Min Assumptions Tested	Min Total Findings
1x	4	2	1	15
2x	6	3	2	28
4x	8	5	4	45
8x	12	7	6	75

Default: 2x. These are floors.

---

Anti-Failure Checks

Failure Mode	Signal	Fix
Taking the want at face value	Jumping straight to methods for the stated want	Trace implications first. The actual want emerges from the commitment chain.
No foreclosures	Only listing what the want opens up	Every want also forecloses. Find what becomes impossible.
Skipping prerequisites	"Here's how to do it" without checking readiness	Every want has prerequisites. Surface them.
No do-nothing path	Missing the option of not pursuing the want	Always map what happens if the user does nothing. Sometimes it's fine.
No unconventional path	Every path is the obvious approach	At least one path should reframe the problem.
Crux-free	No crux identified	If there's no crux, either the want is trivial or you haven't dug deep enough.
All paths viable	Every path looks good	Find the costs and foreclosures. Every path gives something up.
Examples ignored	G5/G11 contradict the user's concrete examples	If user gave examples like "what's next", "how do I", and you labeled the implicit want as adversarial/stress-testing — the examples override abstract framing. Re-read the examples and re-derive.

---

Pre-Completion Check

• Want unbundled into desire, method, belief, assumption, implicit want, anti-want
• Implications traced (assume right — what does the want commit you to?)
• Actual want identified via implication chain (distinguished from stated want)
• Foreclosures identified (what the want makes impossible)
• Prerequisites surfaced with met/unmet/unknown status
• Blockers identified with removable/not assessment
• At least [depth minimum] paths mapped including do-nothing and unconventional
• Crux identified (or explicitly stated as absent with reasoning)
• ALL findings from Phase 1 in registry (none dropped)
• Synthesis introduces NO new findings
• First actions are specific and assigned (Claude or user)
• Output routes to appropriate next skill
• Example check: If user provided concrete examples, every G-item is consistent with them. Examples > abstract framing.