ablation-planner

Generate an Ablation Plan with component changes and hyperparameter sweeps.

1|Updated Mar 24, 2026
One-click install
npx skills add https://github.com/kitcaf/skills --skill ablation-planner-kitcaf
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ablation-planner
Source: https://github.com/kitcaf/skills/tree/main/skills/skills-codex/skills/ablation-planner
Command: npx skills add https://github.com/kitcaf/skills --skill ablation-planner-kitcaf

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Ablation planning is a time-consuming, error-prone task in academic workflows; this skill designs ablations from a reviewer perspective and coordinates between planner and executor to produce a robust evidence plan.

Core Features & Use Cases

  • Systematically design ablations that address reviewer questions and strengthen manuscript submissions.
  • Delegate the design to a secondary Codex reviewer agent while the local executor assesses feasibility and implements.
  • Generate a structured Ablation Plan including component removals/replacements, hyperparameter sweeps, and estimated compute, ready for recording in EXPERIMENT_LOG.md.

Quick Start

Provide project context and the current result-to-claim status to generate a prioritized ablation plan.

Frequently Asked Questions about ablation-planner

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

FAQPage Schema
How do I design ablation studies that address reviewer questions for machine learning manuscripts?

Designing ablations from a reviewer perspective involves coordinating a secondary Codex reviewer agent with a local executor to assess feasibility and implement changes. It produces a structured Ablation Plan with component modifications, hyperparameter explorations, and estimated compute for manuscript submission.

What is the best way to plan machine learning experimental ablations for manuscript submission?

The best way to plan experimental ablations is applying this skill when main results pass claim-supported criteria. It guides reviewer-focused designs and feasibility checks between a reviewer agent and local executor, outputting an Ablation Plan suitable for recording in EXPERIMENT_LOG.md.

Do I need main results to pass claim-supported criteria before planning ablations?

Yes, you need main results to pass claim-supported criteria before planning ablations. The skill applies specifically when this status is met, guiding reviewer-focused experimental designs and feasibility checks to validate the claims your manuscript makes.

Can I use a reviewer agent to check ablation feasibility while a local executor implements it?

Yes, delegating ablation design to a secondary Codex reviewer agent while the local executor assesses feasibility and implements changes is supported. This coordination produces a structured Ablation Plan with component-level changes, hyperparameter explorations, and estimated compute.

What components should an ablation plan include for machine learning research?

An ablation plan should include component-level changes like removals or replacements, hyperparameter explorations, and estimated compute. It generates a structured record suitable for EXPERIMENT_LOG.md to validate claims and strengthen your manuscript submission against reviewer questions.