# Explore Code

> Rigor Improve implementation leaf skill for auditable candidate implementation in deep learning research repositories. Use when the researcher explicitly authorizes exploratory work on an isolated branch or worktree to transplant modules, adapt a backbone, add LoRA or adapter layers, replace a head, or stitch together meaningful low-risk migration ideas with rollback-aware records in `explore_outputs/`. Do not use for end-to-end exploration orchestration on top of `current_research`, trusted baseline reproduction, conservative debugging, environment setup, verified contribution claims, or default repository analysis.

Source: https://skillsagentes.com/skills/lllllllama/rigorpilot-skills/explore-code
Repository: https://github.com/lllllllama/RigorPilot-Skills
Author: lllllllama
License: MIT
Updated: hace 19 días
Context cost: 156 tok installed, 637 tok once triggered, 4.7k tok with every bundled file
Bundle: 5 files, 18 KB
Permissions requested: none declared

## Install

```bash
npx -y skills add lllllllama/RigorPilot-Skills --skill explore-code --agent claude-code
```

## What it does

- Realiza modificaciones exploratorias de código en una rama o worktree aislada
- Transplanta módulos, adapta backbones, inserta capas LoRA/adapter o reemplaza cabezas
- Registra el cambio candidato con motivo, forma de revertirlo y por qué no es aún una contribución verificada
- Genera archivos en explore_outputs/: CHANGESET.md, SCIENTIFIC_CHANGELOG.md, COMPARABILITY_REPORT.md, TOP_RUNS.md, status.json

## Use it when

- El investigador autoriza explícitamente cambios de código exploratorios en una rama o worktree aislada
- La tarea es transplante de módulo, adaptación de backbone, inserción de LoRA/adapter o combinación de módulos de bajo riesgo
- Basta un registro a nivel resumen porque el resultado es un candidato, no una conclusión confiable

## Don't bother when

- Trabajo en la línea base confiable, depuración conservadora o ejecución normal de entrenamiento
- El usuario no autorizó explícitamente modificaciones exploratorias
- La tarea es un refactor amplio o una implementación de idea desde cero

## What triggers it

- "Autorizo probar insertar capas LoRA en el backbone en una rama aislada"
- "Transplanta este módulo del otro repo y documenta el candidato con plan de rollback"
- "Adapta el backbone actual con un adapter y registra por qué es solo candidato"

## Before you install

- Requiere autorización explícita del investigador y trabajar sobre una rama o worktree aislada, además de los scripts plan_code_changes.py y write_outputs.py.

## Files

- SKILL.md — 2 KB
- agents/openai.yaml — 326 B
- references/explore-policy.md — 652 B
- scripts/plan_code_changes.py — 14 KB
- scripts/write_outputs.py — 822 B

## SKILL.md

Reproduced verbatim from lllllllama/RigorPilot-Skills under MIT. This section is the upstream document and is in English.

# explore-code

Use this as the Rigor Improve implementation leaf skill. The installed slug
remains `explore-code` for compatibility.

Use the shared operating principles in
`../../references/agent-operating-principles.md`; this skill should guide
bounded candidate code work without over-prescribing implementation details.

## When to apply

- When the researcher explicitly authorizes exploratory code changes on an isolated branch or worktree.
- When the task is source-anchored module transplant, backbone adaptation, LoRA or adapter insertion, or low-risk module combination.
- When summary-level recording is sufficient and the result is a candidate, not a trusted conclusion.

## When not to apply

- When the request is for trusted baseline work, conservative debugging, or normal training execution.
- When the user did not explicitly authorize exploratory modifications.
- When the task is a broad refactor or a from-scratch idea implementation.

## Clear boundaries

- This skill owns exploratory code modifications only.
- It must keep work isolated from the trusted baseline.
- Use `ai-research-explore` instead when the task spans both current_research coordination and exploratory runs.
- It may hand off execution to `minimal-run-and-audit` or `run-train`.
- It should favor source-anchored copying and minimal adaptation over freeform rewrites.
- It should record why a candidate change is meaningful, how to roll it back,
  and why it remains a candidate rather than a verified contribution.

## Output expectations

- `explore_outputs/CHANGESET.md`
- `explore_outputs/SCIENTIFIC_CHANGELOG.md`
- `explore_outputs/COMPARABILITY_REPORT.md`
- `explore_outputs/TOP_RUNS.md`
- `explore_outputs/status.json`

## Notes

Use `references/explore-policy.md`, `../../references/research-rigor-principles.md`, `scripts/plan_code_changes.py`, and `scripts/write_outputs.py`.

---

Skills Agentes — https://skillsagentes.com/skills/lllllllama/rigorpilot-skills/explore-code