Ai Research Reproduction
Orchestrate README-first, minimal-trustworthy reproduction of a deep learning repository with auditable repro_outputs evidence.
Overview
ai-research-reproduction is an agent skill most often used in Build (also Validate, Ship) that orchestrates README-first deep learning repository reproduction with a minimal trustworthy run and repro_outputs evidence bun
Install
npx skills add https://github.com/lllllllama/rigorpilot-skills --skill ai-research-reproductionWhat is this skill?
- README-first orchestrator for deep learning repository reproduction
- Selects smallest documented inference or evaluation target before expanding scope
- Coordinates intake, setup, trusted execution, optional training and analysis
- Enforces conservative patch rules and records assumptions, deviations, and human decision points
- Writes standardized repro_outputs/ bundle—not score chasing or silent protocol changes
Adoption & trust: 32.3k installs on skills.sh; 412 GitHub stars; 3/3 security scanners passed (skills.sh audits).
What problem does it solve?
You have a DL research repo and need a faithful, auditable reproduction path instead of ad-hoc environment hacks or undocumented protocol changes.
Who is it for?
End-to-end, repository-grounded reproduction when the goal is minimal trustworthy execution aligned with README and recorded evidence.
Skip if: Paper summaries only, generic environment setup without repro intent, silent command execution, score chasing, or research assistance detached from the repo’s documented flow.
When should I use this skill?
User wants an end-to-end, minimal-trustworthy README-first deep learning repository reproduction with auditable evidence.
What do I get? / Deliverables
You get a documented reproduction attempt with repro_outputs artifacts, recorded deviations and decision points, and clear hooks to paper-context-resolver or explore-code when gaps or authorized experiments arise.
- repro_outputs/ standardized bundle
- Recorded assumptions, deviations, and decision points
Recommended Skills
Journey fit
Spans multiple journey phases - primary shelf plus alternate fits below.
Repository-grounded DL reproduction is primarily a build activity: wiring environment, weights, datasets, and documented commands into a runnable, evidenced outcome. integrations fits because the orchestrator coordinates intake, setup, execution, and optional helpers across repo, env, and artifact boundaries rather than a single frontend or docs task.
Where it fits
Prove the published repo runs the smallest documented eval before committing to a full product build on top of the model.
Walk intake through setup and trusted execution while logging dataset and weight assumptions in repro_outputs.
Re-run documented inference commands to verify the baseline still matches README after dependency updates.
How it compares
Orchestrated reproduction workflow with evidence bundles—not a one-off terminal command skill or general literature review.
Common Questions / FAQ
Who is ai-research-reproduction for?
Indie builders and researchers reproducing published DL code who want agent-guided, README-faithful runs with written evidence rather than opaque chat debugging.
When should I use ai-research-reproduction?
When validating a repo’s real runnable surface in validate/prototype, executing the build-phase repro pipeline, or ship-phase verification that documented eval commands still pass—with optional paper-context-resolver for narrow gaps.
Is ai-research-reproduction safe to install?
It is designed for conservative patches and explicit human decision points; review the Security Audits panel on this page and expect shell, network, and filesystem use when running repo commands.
Workflow Chain
Then invoke: paper context resolver, explore code
SKILL.md
READMESKILL.md - Ai Research Reproduction
# ai-research-reproduction ## Purpose Use this as the RigorPilot Reproduce orchestrator for README-first deep learning repository reproduction. The skill guides the agent toward a minimal trustworthy run with auditable evidence; it should not micromanage implementation details that the model can infer from the repository. Reproduction is not "make it run by changing anything"; it means faithfully reading the README, environment, weights, datasets, and documented commands, then recording results and deviations. Start from the shared operating principles in `../../references/agent-operating-principles.md`, then load `../../references/research-rigor-principles.md` and `../../references/deep-learning-experiment-principles.md` when scientific meaning, comparability, or experiment details are at stake. ## Fit Use this skill when all are true: - The target is an AI code repository with a README, scripts, configs, or documented commands. - The request spans multiple trusted phases such as intake, setup, execution, training verification, analysis, paper-gap resolution, and reporting. - The desired result is a small reproducible target, not broad experimentation. Do not use this skill for paper summaries, generic environment setup, isolated repo scanning, standalone command execution, open-ended research design, or explicit candidate-only 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. Read the README and nearby repo signals. 2. Use `repo-intake-and-plan` to extract documented commands and candidate targets. 3. Select and justify the minimum trustworthy target. 4. Use `env-and-assets-bootstrap` only for target-specific environment, checkpoint, dataset, and cache assumptions. 5. Use `analyze-project` only when structure, insertion points, or suspicious implementation patterns need read-only clarification. 6. Use `minimal-run-and-audit` for documented inference, evaluation, smoke, or sanity execution. 7. Use `run-train` instead when the selected trusted target is training startup, short-run verification, full kickoff, or resume. 8. Pause for human review before fuller training claims or any change that could alter dataset, split, checkpoint, preprocessing, metric, loss, model semantics, or result interpretation. 9. Write the standardized outputs and give a concise final note in the user's language when practical. ## 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