agentleFS
Sign inStart free
wizardpc-comwizardpc-com/autonomous-math-ai/.agents/skills/math-research/SKILL.md

math-research

Skill2 starsChanged 3 months agoMITView on GitHub

Conduct research-grade mathematics with cheap falsification, exact or symbolic computation, proof or counterexample search, reproducible evidence, independent audit, or selective formalization. Use for nontrivial claims needing rigorous verification; do not use for routine calculations or explanations.

4 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions4 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 4

    1. sjtuytc/ResearchMathAgent★ 35same name
    2. jelbirt/LLM-Skill-Build★ 1same name
    3. jmf-enigma/codex-theory-proof-workbench★ 1same name
    4. jmf-enigma/math-research★ 1same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repositoryOne other file it gives its agents.

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Prefer the cheapest exact falsification that can meaningfully test the claim.

    line 19
  2. Very specific:

    Every target project declares its own portable paths in autonomous/project.json.

    line 60
  3. Distinctive:

    Create a project-local experiment record with:

    line 65

“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Math Research

research controller decides -> local backend computes -> deterministic verifier judges -> artifacts persist

Research controller

  • Preserve the exact statement, domain, quantifiers, assumptions, and requested mode. Strategy, lemma choice, interpretation, and escalation stay with the capable parent model.
  • Read compute-orchestration.md before parallel work and falsification-policy.md before expensive proof work.
  • Prefer the cheapest exact falsification that can meaningfully test the claim. Use tool-routing.md and local-compute.md to choose local CPU, CAS, SAT/SMT, GPU, or formal tools.
  • Assign evidence only through verification-levels.md. Failed search and raw CAS output are not proof. Read formalization-policy.md before substantial formalization.
  • Refresh a target project's local tool inventory with amr detect-tools --project-root <path>; the inventory is project state, not harness source.

Controlled one-shot mechanical worker

Delegate only a bounded, repetitive task with an exact objective, fixed inputs, finite bounds, a stop condition, and mechanically checkable output. Never delegate proof strategy, invariant choice, research direction, or a final audit verdict. Roles may request a task, but only the persistent controller may start the worker.

Read worker-contract.md and invoke the controller-owned broker:

amr mechanical-run path/to/task.json --project-root path/to/project --timeout 600

The fixed primary route is gpt-5.3-codex-spark / high / service_tier=null. Only an explicit permanent unavailable/access-denied result for that exact route may fall back once to gpt-5.6-luna / medium / service_tier=null. Transient failures receive finite same-route retries. Never use Sol, Terra, the parent model, fast, priority, or a third fallback for the mechanical worker.

The child is one-shot, isolated, network-disabled, and unable to spawn another agent. Its result is mechanical evidence only; it cannot change canonical claims, proofs, trusted state, or an audit verdict.

Project evidence

Every target project declares its own portable paths in autonomous/project.json. Keep all statements, prompts, claims, tasks, experiments, runs, outcomes, audits, and artifacts under that target project, never in this harness repository.

Create a project-local experiment record with:

amr new-experiment <problem-id> --project-root <path> --statement "<exact statement>"

Follow experiment-protocol.md and project-state.md.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free