probe

name	probe
description	Diagnose thinking failures; audit whether reasoning serves inquiry or defense. Use when reasoning feels stuck or circular, a conclusion feels defended not discovered, confidence is high but evidence thin, analysis grows elaborate without growing accurate, or the same approach keeps failing. Trigger phrases: "check my thinking", "am I reasoning well", "why am I stuck", "reasoning feels circular", "probe my logic", "/empire-product:probe". Two modes: self-monitoring (audit own process) and user coaching (diagnose user's thinking with questions, not declarations).
compatibility	Designed for Claude Code (or similar agents).

Good thinking is an active achievement. Operations without orientation produce sophisticated wrong answers. Orientation without operations produces good intentions with no traction.

Two types must work together:

• Operations — cognitive verbs that transform representations (Decouple, Differentiate, Match, Monitor, Hold, Compress, and their complements). Agnostic about what they serve. Same operations produce insight OR sophisticated self-deception.
• Orientations — what operations are in service of. Defined by what stays fixed while everything else moves.

Process-sovereignty (target orientation): inquiry is responsive to evidence. Conclusions move when evidence demands. What is fixed = the meta-commitment to responsiveness itself.

Captured orientations (what to diagnose):

• Conclusion-preservation — a conclusion is fixed; the process bends to defend it
• Authority-preservation — being the authority is fixed; conclusions flex to maintain status
• Threat-reduction — discomfort drives resolution; complexity misread as danger
• Completion-seeking — producing output that sounds good is the goal, not accuracy

SHOULD read CONTEXT.md at repo root if present — use its vocabulary verbatim; imprecise terminology can mask reasoning failures. If CONTEXT.md does not exist, proceed without it. Do not create it unprompted.

Seven operations in complementary pairs. Skill = oscillation between poles, not picking one.

Operation	Action	Failure Modes
Decouple / Re-couple	Detach from binding (belief, identity, context); reattach to reality	Fusion (can't separate idea from frame). Dissociation (detach without grounding back).
Differentiate / Integrate	Split: find joints. Merge: find common structure.	Under-differentiation. Analysis paralysis. Premature integration. "And also" listing without connecting.
Match	Detect structural correspondence; anomaly detection ("this should be like that but isn't")	False matching. Surface matching (visible features not deep structure). Restricted match space from premature framing.
Monitor / Interrupt	Evaluate whether current process is working; intervene when not. Always active.	Co-option: under captured orientation, Monitor actively defends the wrong orientation. The self-corrective machinery becomes self-protective.
Hold / Resolve	Hold: maintain unresolved tension against collapse pressure. Resolve: commit.	Premature resolution (narrows recognition space). Perpetual hold (avoidance disguised as open-mindedness).
Compress / Expand	Build a map from territory, discarding detail. Return to source, recover what was discarded.	Compression that imposes structure source lacks. Map mistaken for territory.

State	Name	What's Fixed	Mechanism	First Move
GT0	No Orientation Awareness	Nothing (no metacognition)	Inertial	Introduce orientation concept
GT1	Conclusion-Preservation	A specific conclusion	Identity fusion	Decouple conclusion from identity
GT2	Authority-Preservation	Being the authority	Identity fusion	Differentiate authority from accuracy
GT3	Threat-Reduction	Comfort / safety	State activation	Address the state, then re-evaluate
GT4	Completion-Seeking	Producing output	Inertial / state	Hold before Resolve
GT5	Monitor Co-option	The defense itself	Identity fusion	External Monitor scaffolding
GT6	Operation Imbalance	One operation pole	Inertial	Deploy the neglected pole
GT7	Premature Resolution	The first frame	Inertial / state	Re-open Hold, generate alternatives

GT5 warning: Cannot be fixed by more monitoring — that's the trap. More metacognitive activity that produces refinements but never reversals is the signature. Introduce external scaffolding (prediction tracking, outside feedback, literal scorekeeping).

GT7 warning: Early frame-commitment feels like clarity, not narrowing. Satisfaction-of-search: having found one answer, fails to recognize others even when looking directly at them.

1. Identify what's fixed. Ask: "In this thinking process, what is not moving — what conclusion, role, comfort level, or output goal is treated as the immovable point?" If nothing is fixed except commitment to responsive inquiry, process-sovereignty is active.

2. Check Monitor. Is Monitor serving inquiry or defending a position? Red flag: metacognitive activity produces refinements but never reversals; engagement with counter-evidence always ends at the same conclusion.

3. Assess operation balance. Which operation pairs are active? Which poles are collapsed? Is the balance matched to the environment's structure, or habitual? Particular attention: is Hold active or collapsed? Is Decouple available or blocked by fusion?

4. Match the capture mechanism. Three mechanisms require different interventions:

   • Identity fusion — conclusion/role/method is part of self-concept. Monitor co-opted. Cannot be fixed by "just think harder." Requires Decouple applied to the identity binding + external Monitor scaffolding.
   • State activation — physiological state shifted orientation pre-deliberatively. Address the state first (safety, resources, time), then re-evaluate orientation.
   • Inertial — no active defense. Orientation appropriate at one point continues without re-evaluation. A prompt to re-evaluate is often sufficient.

5. Select intervention based on mechanism, not surface behavior. Same surface error (e.g., anchoring) can have different mechanisms. A simple re-evaluation prompt works for inertial capture and gets defended against in identity fusion.

Self-monitoring mode — apply to own process. Run diagnostic at key decision points. Watch for: elaboration that defends rather than discovers; confidence without proportionate analysis; restricted match space from premature framing; Monitor co-option producing thorough-feeling but uncorrecting analysis.

User coaching mode — diagnose the user's thinking pattern; guide with questions, not declarations. Lead with Key Questions from the relevant state. Goal: restore process-sovereignty, not impose conclusions about their thinking.

Coaching sequence:

1. Listen for symptoms — which state(s) might be active?
2. Ask diagnostic questions — confirm the mechanism
3. Name the pattern: "Here's what I'm noticing..."
4. Offer the relevant operation: "What if we tried [specific operation] here?"
5. Track whether intervention produces actual movement or better defense

Universal diagnostic questions — start here before state-specific:

1. What is fixed? "What conclusion, role, comfort level, or output goal is treated as the immovable point?"
2. Is Monitor serving inquiry or defending a position? "When did metacognitive checking last produce an actual course correction, not just a refinement?"
3. Which operations are active? "Which poles are collapsed?"
4. What is the mechanism? "Inertial, identity fusion, or state activation?"
5. Does the intervention match the mechanism? "Would a simple re-evaluation prompt work here, or be defended against?"

The Bias Checklist — treating good thinking as error avoidance; running through known biases and checking them off. Operates at surface-behavior level; cannot distinguish underlying mechanisms. Fix: diagnose orientation first.

The Sophistication Trap — responding to doubt about reasoning quality with more analysis, more careful reasoning, without checking orientation. More powerful operations under wrong orientation produce more effective defense of wrong conclusions. Fix: check orientation before adding analytical depth.

The Uniform Fix — applying the same intervention regardless of capture mechanism. A prompt to re-evaluate works for inertial capture and gets actively defended against in identity fusion. Fix: always diagnose the mechanism before selecting an intervention.

The Perpetual Hold — never resolving, never committing, disguised as open-mindedness. Hold and Resolve are a complementary pair. Fix: apply Monitor to the holding itself — is continued holding serving inquiry or serving comfort?

The Surface Diagnosis — matching observed behavior to a state label without investigating the underlying mechanism. Fix: go one level deeper; identify the mechanism.