Calibrate

Calibrate -- Teach Your Clone How You Think

Interactive scenario-based calibration sessions that extract your decision heuristics, values, and reasoning patterns. Each session takes 5-10 minutes and significantly improves how well the clone models your thinking.

How It Works

1. Gap Analysis -- Check which domains of your thinking are under-modeled
2. Scenario Presentation -- Present a realistic scenario that forces a judgment call
3. Adaptive Follow-ups -- Probe deeper based on your initial response
4. Disagreement Probes -- Challenge your answer to surface nuance and exceptions
5. Knowledge Storage -- Extract and store decision patterns and values at high confidence

Backend tools (use these, don't improvise)

• calibration_status (mcp__nomos-think__calibration_status) -- the real

gap analysis (per-domain coverage from the documented patterns*0.6 + values*0.4 formula). Use for /calibrate status and to pick the focus domain.

• calibration_next_scenario (mcp__nomos-think__calibration_next_scenario)

-- returns the next scenario (from the curated library) targeting the biggest gap. Pass completedIds for scenarios already shown this session. Present the scenario it returns; do not invent your own.

Commands

• /calibrate -- Start a calibration session (auto-picks the least-covered domain)
• /calibrate status -- Show calibration coverage and gaps
• /calibrate -- Focus on a specific domain

Available Domains

techdecisions, communication, conflict, prioritization, leadership, quality, collaboration, risk, creativity, timemanagement

Session Protocol

When the user invokes /calibrate, follow this exact protocol:

Phase 1: Status Check

1. Call user_model_recall to load the current user model
2. Analyze coverage across all calibration domains
3. Show a brief status summary:

``` Clone calibration: 42% [========------------]

Gaps: Tech Decisions (10%), Prioritization (0%), Risk (20%) Best covered: Communication (80%), Quality (70%)

Starting session on: Prioritization (least covered) ```

Phase 2: Scenario

1. Present ONE scenario from the target domain
2. Frame it conversationally: "Here's a situation I'd like to understand how you'd handle..."
3. Wait for the user's response -- do NOT rush or provide options

Phase 3: Follow-up Probes

After the user responds, ask 2-3 follow-up questions:

1. Clarification: "What's the main factor driving that choice?"
2. Exception probe: "Would anything flip your decision? What would have to be different?"
3. Disagreement probe: "I think based on what I know about you, you'd lean toward [opposite]. Am I wrong?" (Only use this if you have a genuine hypothesis from the user model)

Phase 4: Extraction & Storage

After the probing conversation, extract the knowledge:

1. Identify the decision pattern (the heuristic or principle behind their choice)
2. Identify any values revealed (what they prioritize and why)
3. Identify any exceptions to the pattern

Store these using user_model_recall confirmation -- tell the user what you learned:

Here's what I learned from this scenario:

Decision pattern: "Ship the working solution first, refactor later -- unless
tech debt would block other teams" (weight: 0.8)
  Context: deadline pressure, stakeholder management
  Exception: when the debt affects shared infrastructure

Value: Pragmatism over perfectionism in time-constrained situations

Does this capture it accurately? Anything to adjust?

Phase 5: User Correction

If the user corrects or refines your extraction:

• Update the pattern/value immediately
• Store with confidence 0.85 (explicit calibration = high confidence)
• Thank them and note what you adjusted

Phase 6: Continue or Close

Ask: "Want to continue with another scenario, or is this a good stopping point?"

If continuing, pick the next least-covered domain.

Important Rules

• One scenario at a time -- never present multiple scenarios
• Wait for responses -- don't anticipate or provide sample answers
• Be genuinely curious -- these are real conversations, not quizzes
• Store at high confidence (0.85) -- explicit calibration is the most reliable signal
• Show what you learned -- always summarize extractions and ask for confirmation
• Track progress -- update the coverage % as you go
• Respect time -- if the user seems done, gracefully close even if there are more gaps
• No judgment -- there are no right or wrong answers, only preferences to understand

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: project-nomos
• Source: project-nomos/nomos
• License: MIT
• Homepage: https://projectnomos.dev

Install and usage instructions live in the source repository linked above.