# Ai Research Reproduction

> Skill compatible con Rigor Reproduce para reproducción README-first de repos de deep learning: elige el objetivo mínimo confiable, coordina fases confiables y registra evidencia y desviaciones en `repro_outputs/`.

Fuente: https://skillsagentes.com/skills/lllllllama/rigorpilot-skills/ai-research-reproduction
Markdown: https://skillsagentes.com/skills/lllllllama/rigorpilot-skills/ai-research-reproduction.md

Repositorio: https://github.com/lllllllama/RigorPilot-Skills
Autor: lllllllama
Licencia: MIT
Actualizado: hace 2 meses
Coste de contexto: 192 tok instalada, 1.7k tok al activarse, 20.6k tok con todos los archivos del bundle
Bundle: 14 archivos, 80 KB
Permisos que pide: ninguno declarado

## Instalación

Un skill son archivos markdown: los mismos archivos valen para cualquier agente y lo único que cambia es el directorio de destino, es decir la bandera `--agent`. Añade `-g` para instalarlo en todos los proyectos de la máquina.

```bash
# Claude Code
npx -y skills add lllllllama/RigorPilot-Skills --skill ai-research-reproduction --agent claude-code
# Cursor
npx -y skills add lllllllama/RigorPilot-Skills --skill ai-research-reproduction --agent cursor
# Codex
npx -y skills add lllllllama/RigorPilot-Skills --skill ai-research-reproduction --agent codex
# Gemini CLI
npx -y skills add lllllllama/RigorPilot-Skills --skill ai-research-reproduction --agent gemini
# Windsurf
npx -y skills add lllllllama/RigorPilot-Skills --skill ai-research-reproduction --agent windsurf
# Cline
npx -y skills add lllllllama/RigorPilot-Skills --skill ai-research-reproduction --agent cline
```

## Qué hace

- Coordina un flujo README-first para reproducir repositorios de deep learning: intake, setup, ejecución confiable, entrenamiento opcional, análisis opcional y resolución de brechas con el paper
- Selecciona el objetivo mínimo demostrable (inferencia, evaluación o arranque de entrenamiento documentados) en lugar de ejecutar todo
- Aplica reglas de parches conservadoras y exige registrar evidencia, supuestos, desviaciones y puntos de decisión humana
- Escribe el paquete estandarizado en `repro_outputs/` con SUMMARY.md, COMMANDS.md, LOG.md, SCIENTIFIC_CHANGELOG.md, COMPARABILITY_REPORT.md, status.json y ANNOTATED_README.md
- Pausa para revisión humana antes de afirmaciones de entrenamiento más completas o cambios que alteren dataset, split, checkpoint, métrica o semántica del modelo

## Cuándo usarla

- El objetivo es un repositorio de código de IA con README, scripts, configs o comandos documentados
- La solicitud abarca varias fases confiables: intake, setup, ejecución, verificación de entrenamiento, análisis, resolución de brechas con el paper y reporte
- El resultado deseado es un objetivo pequeño y reproducible, no experimentación amplia

## Cuándo no

- Resúmenes de papers
- Configuración genérica de entorno o escaneo aislado de un repo
- Ejecución de comandos independientes, diseño de investigación abierto o exploración solo de candidatos

## Qué la activa

- "Reproduce el resultado de inferencia documentado en este repositorio siguiendo el README"
- "Ejecuta la evaluación mínima documentada de este repo de deep learning y registra la evidencia"
- "Necesito reproducir de forma confiable el arranque de entrenamiento de este proyecto y documentar las desviaciones"

## Antes de instalar

- Requiere un repositorio con README, scripts o comandos documentados como base para la reproducción.
- makes network requests

## Archivos

- SKILL.md — 6 KB
- agents/openai.yaml — 389 B
- assets/COMMANDS.template.md — 348 B
- assets/LOG.template.md — 669 B
- assets/PATCHES.template.md — 525 B
- assets/SUMMARY.template.md — 621 B
- assets/status.template.json — 1 KB
- references/architecture.md — 1 KB
- references/language-policy.md — 794 B
- references/output-spec.md — 3 KB
- references/patch-policy.md — 2 KB
- references/research-safety-principles.md — 2 KB
- scripts/annotate_readme.py — 19 KB
- scripts/orchestrate_repro.py — 43 KB

## SKILL.md

Reproducido tal cual desde lllllllama/RigorPilot-Skills bajo MIT. Esta sección es el documento original y está en inglés.

# ai-research-reproduction

## Purpose

Guide README-first deep learning reproduction toward the smallest trustworthy
run with auditable evidence. Preserve documented meaning; record assumptions,
deviations and blockers instead of changing semantics to manufacture success.
Load specialized references only for a concrete uncertainty.

## Fast Path

For a routine bounded run, keep the control path short:

1. Read the target README and only the target test/config/source needed to understand the documented command.
2. Run `scripts/orchestrate_repro.py --repo <repo> --plan-only --agent-output` with any explicit user timeout bound (`--timeout` or `--train-timeout`) already supplied; review `command_candidates`, the selected `cmd-XX`, side-effect contract, selection fingerprint, and returned `reviewed_run_args`. With no `--output-dir`, later evidence goes to `<repo>/repro_outputs` regardless of caller cwd.
3. Run the selected candidate, or another reviewed candidate, with `--run-selected --command-id <cmd-XX> --plan-fingerprint <fingerprint> --agent-output` plus requested timeout/metric/source-adjacent options. Preserve an explicit user command-timeout bound instead of silently making it stricter on a routine trusted run. `--timeout` limits the target command; do **not** wrap the whole orchestrator in an equal or shorter external timeout, because it still needs time to terminate children and write terminal evidence. A changed command set fails closed; setup/download commands are never target candidates.
4. Run `--verify-output --agent-output`; inspect detailed evidence files only when verification fails or the result is partial/blocked.
5. Deliver the bounded result and stop.

For a host with short tool-call deadlines, rerun planning with `--include-agent-handoff` and follow `references/agent-job.md`; otherwise keep the direct path above. Job completion is not task acceptance, and uncertain state is never a reason for automatic replay.

Do **not** inspect `orchestrate_repro.py`, `annotate_readme.py`, `_bundled/`, writers, or runtime internals on a normal success path. Inspect implementation only for a concrete blocker, unexpected side effect, bundle-integrity failure, or unresolved safety question. Use `scripts/doctor.py` for first-use environment/install diagnostics. Executed commands keep full lifecycle/log evidence under `repro_outputs/_runtime/<run_id>/`.

## Fit

Use this skill for repository-grounded, multi-phase trusted reproduction where
the goal is a small reproducible target. Do not use it for paper summaries,
generic setup, isolated scanning, standalone commands, open-ended research design,
or explicitly authorized candidate exploration.

## Trusted Target Selection

Choose the smallest target that can honestly demonstrate repository-grounded
reproduction:

1. documented inference
2. documented evaluation
3. documented training startup or partial verification
4. full training only after explicit user confirmation

Treat README guidance as the primary reproduction intent. Use repository files
to clarify the README, not to silently replace it. When the README and paper
conflict, record the conflict and use `paper-context-resolver` only for the
narrow reproduction-critical gap.

## Workflow

1. Treat README guidance as primary; extract and select the minimum trustworthy target.
2. Use setup/assets only for target-specific prerequisites and `analyze-project` only when structural clarification is needed.
3. Use `minimal-run-and-audit` for inference/evaluation/smoke and `run-train` for training startup, kickoff, or resume; direct execution is the default.
4. Pause before fuller training or changes to dataset, split, checkpoint, preprocessing, metric, loss, model semantics, or interpretation.
5. Award `result-match` only against explicit expected metrics and tolerance; process success alone is not reproduction success.
6. Write the evidence bundle, return the requested bounded result, and stop; optional stages are not automatic follow-up work.

## Patch Boundary

Prefer no repository edits. If edits are needed, keep them conservative and
auditable:

- Try command-line arguments, environment variables, path fixes, dependency
  version fixes, or dependency-file fixes before code changes.
- Reproduction fixes are allowed when needed, but they must not be hidden. State
  what changed, why it was necessary, whether it changes scientific meaning,
  and whether it affects comparability with the paper, README, or baseline.
- Avoid changing model architecture, core inference semantics, training logic,
  loss functions, or experiment meaning.
- If repository files must change, create a branch named
  `repro/YYYY-MM-DD-short-task`, keep verified patch commits sparse, and record
  README-fidelity impact in `PATCHES.md`.

See `references/patch-policy.md`.

## Outputs

Always target `repro_outputs/`:
```text
SUMMARY.md
COMMANDS.md
LOG.md
SCIENTIFIC_CHANGELOG.md
COMPARABILITY_REPORT.md
status.json
ANNOTATED_README.md   # original README + colored per-section agent-action annotations
PATCHES.md   # only if patches were applied
```

Use the templates under `assets/` and `references/output-spec.md`. Keep summaries
short, commands copyable, machine state stable, and scientific/comparability
changes explicit. `ANNOTATED_README.md` must preserve the source README byte-for-byte
outside inserted evidence blocks and pass its strip/check round trip. Use
`--source-adjacent-readme` only for an owned `RIGORPILOT_README.md`; never replace
an unrelated file. Distinguish verified facts from inference.

## Reference Loading

- Workflow judgment: `references/agent-operating-principles.md`.
- Human-readable output: `references/language-policy.md`.
- Scientific/comparability judgment: `references/research-rigor-principles.md` and, when experiment details matter, `references/deep-learning-experiment-principles.md`.
- Protocol-sensitive changes: `references/research-safety-principles.md` and `references/patch-policy.md`.
- Personal rigor and lessons are advisory only; keep specialized detail in references/scripts rather than expanding this entrypoint.

## Dónde encaja

- Categoría: [Desarrollo de APIs](https://skillsagentes.com/categorias/desarrollo-apis.md) — Diseña, prueba y documenta APIs HTTP y GraphQL.
- Creador: [lllllllama](https://skillsagentes.com/creators/lllllllama.md) — 11 skills en el directorio
- [Todas las skills](https://skillsagentes.com/skills.md)
- [Ranking de instalaciones](https://skillsagentes.com/ranking.md)

## Otras skills del mismo repositorio

- [Ai Research Explore](https://skillsagentes.com/skills/lllllllama/rigorpilot-skills/ai-research-explore.md): Slug de skill compatible con Rigor Explore para candidatos de investigación en deep learning potencialmente novedosos: exploración solo de candidatos sobre `current_research`, con comprensión auditable del repo y comparación justa.
- [Minimal Run And Audit](https://skillsagentes.com/skills/lllllllama/rigorpilot-skills/minimal-run-and-audit.md): Skill Rigor Run para reproducir repos de deep learning centrados en el README: captura evidencia de un smoke test, inferencia o evaluación documentada en repro_outputs/, con notas de parches si cambian archivos.
- [Repo Intake And Plan](https://skillsagentes.com/skills/lllllllama/rigorpilot-skills/repo-intake-and-plan.md): Ayudante de Rigor Intake para reproducción de repos de deep learning basada en README: escanea el repo, extrae comandos documentados y clasifica candidatos de inferencia, evaluación y entrenamiento.
- [Run Train](https://skillsagentes.com/skills/lllllllama/rigorpilot-skills/run-train.md): Skill de Rigor Train para repositorios de investigación de deep learning: ejecuta de forma conservadora comandos de entrenamiento y registra evidencia estandarizada en train_outputs/.
- [Explore Run](https://skillsagentes.com/skills/lllllllama/rigorpilot-skills/explore-run.md): Skill hoja de Rigor Improve/Rigor Explore para evidencia exploratoria acotada: validaciones en subconjunto, sweeps, búsqueda en GPU ociosa o transfer-learning rápido, con resúmenes sin sobreclamar en `explore_outputs/`.

---

Skills Agentes · [Índice de páginas en markdown](https://skillsagentes.com/sitemap.md) · [Inicio](https://skillsagentes.com/index.md)