# Plan a Trustworthy Repository Reproduction

Repository reproduction often starts without a clear map of documented setup and experiment paths. This skill scans repository signals and recommends a minimal reproduction target.

## Install

```bash
npx skillstore add lllllllama/repo-intake-and-plan
```

## Metadata

- Status: approved
- Slug: lllllllama-repo-intake-and-plan
- Skillstore revision: r2
- Version status: missing
- Tree hash: 4b00aa8074d7eb651716634f86378ee5925e6961d3d448b940b78df58cce2805
- Author: lllllllama
- GitHub username: lllllllama
- License: MIT
- Repository: https://github.com/lllllllama/ai-paper-reproduction-skill/tree/main/skills/repo-intake-and-plan/
- Ref: 656be3040aef5c047555a908cd5c695d22a4a548
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands
- Quality score: 81
- Quality tier: silver
- Public page: https://skillstore.pages.dev/skills/lllllllama-repo-intake-and-plan
- Manifest: https://skillstore.pages.dev/api/skills/lllllllama-repo-intake-and-plan/manifest

## Capabilities

- Scans top-level repository files and high-signal directories.
- Finds README files and common environment or packaging files.
- Extracts shell-like commands from Markdown code blocks and prompts.
- Classifies commands as inference, evaluation, training, or other.
- Marks commands containing placeholders that require user values.
- Recommends a minimal reproduction target while recording ambiguity.

## Use Cases

- Triage a Research Repository: Map documented commands and select a small reproduction target before allocating compute.
- Prepare an Engineering Handoff: Summarize setup files, entrypoints, and unresolved requirements for an implementation team.
- Assess a Student Project: Separate inference, evaluation, and training paths before beginning a course reproduction assignment.

## Prompt Templates

### Basic Repository Intake

```
Scan {repository_path} and summarize its README, important files, documented commands, and smallest trustworthy reproduction target.
```

### Inference-First Plan

```
Inspect {repository_path}, prioritize documented inference commands, identify required substitutions, and recommend the smallest runnable inference path.
```

### Compare Experiment Paths

```
Classify documented commands in {repository_path} as inference, evaluation, training, or other, then compare their evidence and expected complexity.
```

### Conservative Reproduction Handoff

```
Audit {repository_path} for explicit README evidence, distinguish documented facts from inferences, list ambiguities, and produce a minimal plan for the orchestrator.
```

## Limitations

- Does not execute extracted commands or verify their runtime behavior.
- Does not install environments, download assets, or prepare datasets.
- Uses README evidence and filename signals, which can be incomplete.
- Does not perform paper lookup or produce the final reproduction report.

## Best Practices

- Provide the repository root and state whether inference, evaluation, or training has priority.
- Treat extracted commands as evidence for planning, not as approved execution instructions.
- Resolve every reported placeholder and ambiguity before environment setup or execution.

## Anti Patterns

- Do not ask this skill to install dependencies or run experiments.
- Do not treat filename-based classifications as confirmed documentation.
- Do not select expensive training when a documented inference or evaluation target meets the goal.

## Security Audit

- Audited at: 2026-08-08T09:01:19.24\+00:00
- Summary: All four static findings are false positives caused by Markdown backticks, a regular expression, and a supported language label. The scripts read local repository metadata and README content without invoking extracted commands.

## Stats

- Views: 102
- Downloads: 17
- Favorites: 0
- Popularity score: 0