# Analyze Deep Learning Repositories

Research repositories can hide important model, configuration, and evaluation relationships. This skill maps those relationships and reports suspicious patterns without changing project code.

## Install

```bash
npx skillstore add lllllllama/analyze-project
```

## Metadata

- Status: approved
- Slug: lllllllama-analyze-project
- Skillstore revision: r2
- Version status: missing
- Tree hash: 5af961bca6e9d34069cddebf6bead4d8deaa2546b8cdeafa25a15166cebde856
- Author: lllllllama
- GitHub username: lllllllama
- License: MIT
- Repository: https://github.com/lllllllama/rigorpilot-skills/tree/main/skills/analyze-project/
- 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: scripts, external\_commands, filesystem
- Quality score: 81
- Quality tier: silver
- Public page: https://skillstore.pages.dev/skills/lllllllama-analyze-project
- Manifest: https://skillstore.pages.dev/api/skills/lllllllama-analyze-project/manifest

## Capabilities

- Finds likely training, inference, evaluation, model, and configuration entrypoints.
- Maps task-relevant files, data interfaces, output hints, and checkpoint paths.
- Extracts Python class, constructor, forward-method, and configuration-binding hints.
- Builds research, change-zone, and evaluation-contract summaries.
- Flags suspicious activation, dropout, optimizer, and positional-encoding patterns as heuristics.
- Writes structured Markdown summaries and a machine-readable status report.

## Use Cases

- Map an unfamiliar research repository: Identify important entrypoints, model files, configurations, datasets, and outputs before planning changes.
- Review a paper implementation: Create a conservative map of architecture, evaluation logic, and suspicious patterns before reproducing results.
- Prepare a scoped code review: Locate likely change zones and protected evaluation areas before assigning implementation work.

## Prompt Templates

### Basic repository map

```
Analyze this repository and identify its main model, training, inference, evaluation, and configuration files.
```

### Trace model execution

```
Map model construction and forward execution from configuration loading through training and inference entrypoints.
```

### Review evaluation boundaries

```
Identify evaluation commands, metrics, expected artifacts, protected files, and likely risks without running jobs.
```

### Plan a constrained extension

```
Find safe insertion points for [component], list affected interfaces, and separate evidence from heuristic concerns.
```

## Limitations

- Uses filename and text heuristics, so findings require human verification.
- Does not execute training, inference, evaluation, or benchmark commands.
- Does not modify repository source code or implement suggested changes.
- YAML context files require PyYAML to be installed.

## Best Practices

- Provide the repository root and a clear task family.
- Supply evaluation context when metric, dataset, or benchmark details matter.
- Verify every heuristic against the relevant source before planning changes.

## Anti Patterns

- Do not treat heuristic warnings as confirmed bugs.
- Do not use this skill to run heavy training or evaluation jobs.
- Do not request automatic patches or broad speculative refactoring.

## Security Audit

- Audited at: 2026-08-08T08:42:07.566\+00:00
- Summary: Twenty-five static findings are false positives caused by string matching, Markdown backticks, and ordinary Python syntax. Two findings are confirmed because SKILL.md directs agents to read files outside the packaged skill directory.

## Stats

- Views: 0
- Downloads: 23
- Favorites: 1
- Popularity score: 0