reasoning

---

name: reasoning description: Integrated 5-phase reasoning workflow: calibrate state, frame the problem, analyze with lenses, recommend with evidence, verify before acting.

Reasoning

Integrated 5-phase reasoning workflow. Use this when a request needs any combination of state calibration, problem framing, multi-lens analysis, structured advisory, or verification.

This replaces the separate thinking-ground, problem-definition, dynamic-problem-solving, domain-expert-consultation, and self-cognitive skills. Run only the phases the situation needs; skip the rest.

This is a nested child under using-reasoning; its path is using-reasoning/reasoning/.

Core Contract

• Run phases sequentially. Skip any phase the situation does not need.
• Each phase is optional. Enter at the correct phase for the situation.
• Hand off cleanly between phases — output of one is input to the next.
• If a previous phase is still needed, do not jump ahead.
• When the full workflow is done, return the complete output with all relevant phases.

Phase Selection

If the situation is...	Start at
Emotionally charged, looping, defending an answer	Phase 0 — Calibrate
Vague, no clear problem statement, solution-contaminated	Phase 1 — Frame
Clear problem, needs rigorous analysis	Phase 2 — Analyze
Needs structured recommendation or advisory memo	Phase 3 — Recommend
Post-decision, needs verification or pattern extraction	Phase 4 — Verify
Just need to separate facts from assumptions before acting	Phase 0 → Phase 4 (skip 1-3)

---

Phase 0: Calibrate

Improve reasoning quality before analysis. Use when attachment, urgency, looping, or performative analysis is distorting judgment.

State Scan — identify the active distortion using observable signals:

• Defended: same conclusion repeated despite counterevidence; objections explained away; morally loaded language
• Anxious: urgency exceeds the real deadline; catastrophic language; reversible choices treated as irreversible
• Performing: elaborate language without clear next action; complexity grows without understanding
• Open enough to proceed: can state counter-case; new evidence changes answer; uncertainty without panic

Correction protocols — apply the matching protocol:

• Defended → Assumption Strip: name the protected answer; ask "what would have to be true for the opposite to be correct?"
• Anxious → Deceleration: name feared outcome precisely; separate real deadline from emotional pressure; classify as reversible/irreversible
• Performing → Simplicity Test: restate in plain language; remove decorative depth; force one concrete next step
• Open → proceed to Phase 1 or 2

Output: calibrated state label + what changed + next phase

---

Phase 1: Frame

Turn an unclear situation into a single, solution-neutral problem statement.

Phase 0 — Capture situation as lived: Separate observations, interpretations, and attempted fixes.

Phase 1 — Symptom vs root cause: 5-Why drill from visible symptom. Branch on multiple plausible causes. Test each candidate: "If solved, does original discomfort disappear?"

Phase 2 — Reframe five ways:

1. Flip the subject
2. Zoom out (system view)
3. Zoom in (one component)
4. Flip the assumption (wrong goal)
5. Stakeholder swap

Phase 3 — Map boundaries: in scope / out of scope / real constraints / perceived constraints

Phase 4 — Stakeholder reality check: each stakeholder's version of the problem + mirror-imaging check

Phase 5 — Synthesize: [SUBJECT] cannot [BEHAVIOR] because [ROOT CAUSE], which leads to [CONSEQUENCE], despite [ATTEMPTED APPROACH].

Output: one problem statement + boundaries + handoff to Phase 2

---

Phase 2: Analyze

Analyze a clearly defined problem through deliberately chosen lenses. End with tradeoffs and a concrete next action.

Lens selection — choose 2-4 from the lens library (references/lens-library.md):

• Economics lens: incentives, game theory, market structure
• Systems lens: feedback loops, delays, leverage points
• Psychology lens: cognitive biases, motivation, perception
• Engineering lens: tradeoffs, constraints, composability
• Strategy lens: positioning, moats, optionality
• Organizational lens: power, alignment, coordination costs
• Decision lens: framing effects, status quo bias, escalation
• [14 total — see lens library]

Analysis: apply each lens to the problem. Record:

• What each lens reveals
• Where lenses agree and disagree
• What is hidden from each lens

Lens collision protocol: when lenses produce conflicting results:

1. Check which lens's assumptions hold in this context
2. Check if the conflict reveals a hidden tradeoff
3. If unresolvable, escalate to tradeoff assessment

Output: per-lens findings + collision summary + recommendation direction

---

Phase 3: Recommend

Produce a structured recommendation, tradeoff evaluation, or decision memo. Use when the problem is clear and the need is advisory.

Evidence hierarchy: label every claim as:

• Direct evidence: observed, measured, documented
• Strong inference: supported by converging evidence
• Weak inference: plausible but unsupported
• Speculation: no supporting evidence

Recommendation structure:

1. Summary recommendation (one sentence)
2. Options considered (2-5, each with tradeoffs)
3. Chosen option + rationale anchored to evidence
4. Unresolved tradeoffs and their impact
5. Next steps and who should act
6. Conditions that would change the recommendation

Self-check: could an honest person with the same evidence reach a different conclusion? If yes, the recommendation is premature.

Output: decision memo with options, recommendation, evidence, and next steps

---

Phase 4: Verify

Post-execution verification, retrospective, or reusable pattern extraction.

Preflight verification (before acting on the recommendation):

• Separate facts from assumptions
• Confidence calibration: low / medium / high + evidence
• Disconfirming pass: "what evidence is missing that would change the conclusion?"
• Risks and failure modes

Postflight retrospective (after action):

• What changed in system understanding
• Process: what to repeat, stop, or tighten
• Prompting/coordination: what would improve future delegation

Pattern extraction (optional — when reusable knowledge surfaced):

• One concrete, reusable rule or guardrail
• Who should adopt it and under what conditions
• Suggested format: skill patch, checklist, or decision rule

Output: verification statement + optional reusable pattern + confidence

---

Output Format

Return a consolidated response with sections for each executed phase:

Phases Executed

• [list which phases ran: 0-calibrate, 1-frame, 2-analyze, 3-recommend, 4-verify]

[Phase 0 Output — if used]

State: [defended/anxious/performing/open] Correction: [protocol applied + result]

[Phase 1 Output — if used]

Problem statement: [one sentence] Boundaries: [in scope / out of scope]

[Phase 2 Output — if used]

Lenses used: [...] Findings: [per lens] Collision: [where lenses agree/disagree]

[Phase 3 Output — if used]

Recommendation: [one sentence] Options: [2-5 with tradeoffs] Evidence: [hierarchy-labeled]

[Phase 4 Output — if used]

Verification: [facts vs assumptions, confidence] Lessons: [technical / process / coordination] Pattern: [optional — reusable rule or guardrail]

---

References

• references/lens-library.md — 14 analytical lenses with use cases
• references/bias-inventory.md — cognitive bias reference
• references/artifacts.json — optional structured artifact schema
• references/archetypes.md — reality check (survivability audit) template
• references/scenarios.md — scenario planning (strategic foresight) template
• assets/quick-invoke-template.md — quick-start template
• assets/worked-example.md — full worked example

Quick-Start Templates

When the situation maps to one of these patterns, start with the named template instead of running the full 5-phase workflow from scratch.

Reality Check (Survivability Audit)

Best for: career, education, business, or life decisions where hidden rules and survivability matter more than optimization.

1. State the bet — name what the user is staking (money, time, reputation, optionality)
2. Hidden rules — gatekeepers, information asymmetries, selection filters, incentive distortions
3. Survivability — assume optimistic narrative is partly wrong; can they survive the downside?
4. Resource match — compare path requirements against actual position
5. Bottom line — proceed / proceed with conditions / stop

See references/archetypes.md for the full template with quality bar and output format.

Strategic Foresight (Scenario Planning)

Best for: concrete external signals, thresholds, launches, or policy shifts where the user wants implications under uncertainty.

1. Define the signal — what changed and what type of shift is it?
2. Strategic hinge — which assumption may no longer hold?
3. Three scenarios — base case / upside / failure (with time horizons)
4. War-game actors — winners, losers, pivoters
5. Higher-order effects — 1st, 2nd, 3rd order consequences
6. Bottlenecks and risks — what could kill the thesis
7. Decision support — indicators to watch

See references/scenarios.md for the full template with output format and quality bar.

Failure Modes to Avoid

• Running Phase 2 (Analyze) before Phase 1 (Frame) when the problem is vague — solving the wrong problem is still failure
• Running Phase 3 (Recommend) before Phase 2 (Analyze) — recommendations without analysis are opinions
• Running Phase 0 (Calibrate) unnecessarily — don't pathologize normal reasoning
• Confusing Phase 4 (Verify) with Phase 0 (Calibrate) — verify is post-decision, calibrate is pre-analysis
• Jumping to Phase 3 or 4 when Phase 1 has not been done — premature closure is the most common failure mode