Model Assumptions Builder skill

Inputs

• problem parse;
• active method cards;
• data profile and risk-probe summaries;
• question dependency map;
• existing assumptions and human decisions.

Read legacy candidate pools only during migration.

Workflow

1. Extract explicit problem assumptions and method-induced assumptions.
2. Remove filler statements that do not affect model validity or interpretation.
3. For each assumption record:
   • scope and source;
   • modeling need;
   • applicable method/Qx;
   • validation evidence;
   • mitigation or fallback link.
4. Identify conflicts across Qx.
5. Present unresolved necessity/impact trade-offs in one compact choice card where possible.
6. Log human assumption_necessity decisions in qx_decisions.jsonl.
7. Save planning/model_assumptions.md, transcribing settled human labels and impacts with decision IDs.

Assumption Fields

• ID;
• statement;
• scope;
• source and modeling need;
• human-confirmed type: necessary or simplifying;
• validation method/evidence;
• impact if violated;
• mitigation/fallback;
• decision ID.

Rules

• Do not invent generic assumptions such as “data are accurate” unless they affect a real dependency.
• Do not finalize necessary/simplifying or impact judgments for the human.
• Do not leave many repeated sentinels in the final file; collect missing judgments through a choice card and stop finalization until answered.
• Revisit an assumption only when its method, evidence, or downstream use materially changes.

Verification

• Every assumption has a modeling need and source.
• Human-owned labels trace to decisions.
• Probe/robustness evidence addresses load-bearing assumptions.
• Cross-Qx conflicts are resolved or explicit.