idea-formalizer

Formalizes approved research ideas into theorem-ready setting.md artifacts with explicit assumptions and goals.

65|2|Updated Aug 10, 2026
One-click install
npx skills add https://github.com/DechenZhang/VALG-ML-Theory-Agent --skill idea-formalizer-dechenzhang
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: idea-formalizer
Source: https://github.com/DechenZhang/VALG-ML-Theory-Agent/tree/main/skills/idea-formalizer
Command: npx skills add https://github.com/DechenZhang/VALG-ML-Theory-Agent --skill idea-formalizer-dechenzhang

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? It removes ambiguity between idea generation and theorem proving by translating an approved branch-local research idea into a precise mathematical setting and goal, so proof stages can start without redefining the problem or guessing missing assumptions. ## Core Features & Use Cases - Theorem-Ready Formalization: Converts idea.md into a structured setting.md with a Basic setup, Technical assumptions carrying stable assump:<slug> ids, and a single Goal statement. - Failure-Driven Revision Support: Preserves previously valid formalization choices after controller-level idea modification, applying only the approved semantic delta from the recorded failure. - Assumption Discipline: Distinguishes primitive checkable assumptions from derived properties, routing unsupported theorem-critical assumptions back to idea generation instead of adding them silently. - Use Case: In an ML theory research workflow, after a perspective branch's idea is approved, run this skill to produce perspective_M/idea_N/setting.md that downstream proof-sketch stages can consume directly. ## Quick Start Formalize the approved idea in perspective_1/idea_2/idea.md into a theorem-ready setting.md following the shared template.

Frequently Asked Questions about idea-formalizer

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I formalize a research idea into a theorem-ready setting?

Provide the approved idea.md for the branch and run the formalization step, which writes perspective_M/idea_N/setting.md with a Basic setup, Technical assumptions with stable assump:<slug> ids, and one Goal statement. The output follows the shared setting template exactly.

What inputs does the idea formalization step require?

The only required input is perspective_M/idea_N/idea.md. Optional inputs include workflow_feedback.md, LITERATURE_SURVEY.md, prior same-perspective setting.md and proof_review.md files, and any existing setting.md draft for checkpoint reruns.

How does formalization handle failure-driven idea revisions?

It identifies the semantic delta between the prior same-perspective idea and the current approved idea, preserves still-valid formalization choices from the previous setting.md, and rewrites only the modified or invalidated parts. Target-preserving repairs keep the prior theorem target; target-changing repairs are made explicit.

Can the formalizer add new assumptions not in the approved idea?

No. Every theorem-critical assumption must be primitive, checkable, and supported by literature, standard facts, or verification examples recorded in idea.md. If a needed assumption lacks recorded support, the workflow routes back to idea generation instead of adding it silently.

What is the difference between exact-goal mode and target-spec mode?

Exact-goal mode states the final theorem claim precisely when it is already known. Target-spec mode defines the target quantity, claim type, active regime, and success criterion when exact rates or constants are not yet known, without guessing unsupported sharper statements.