# Discover Current AI Models and Tiers

Model identifiers change and quickly make configurations stale. This skill queries supported providers, caches results, detects additions, and recommends capability tiers.

## Install

```bash
npx skillstore add consiliency/model-discovery
```

## Metadata

- Status: approved
- Slug: consiliency-model-discovery
- Skillstore revision: r1
- Version status: missing
- Tree hash: 344ff25920c3b2d3d4744d784c31102429d12f86c103c4921316240483104bcf
- Author: Consiliency
- GitHub username: Consiliency
- License: MIT
- Repository: https://github.com/Consiliency/treesitter-chunker/tree/main/.ai-dev-kit/skills/model-discovery
- Ref: 635f69fb8d2f4e6330ba47a4e5a0fb239c04d110
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: confirmation\_required
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: env\_access, network, filesystem, external\_commands
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/consiliency-model-discovery
- Manifest: https://skillstore.pages.dev/api/skills/consiliency-model-discovery/manifest

## Capabilities

- Fetches model lists from Anthropic, OpenAI, Gemini, and Ollama endpoints.
- Caches provider responses and reuses fresh results for up to 24 hours.
- Detects model identifiers that are absent from the known-model registry.
- Infers fast, default, or heavy tiers from configurable naming patterns.
- Supports JSON reporting, automatic classification, interactive classification, and dry runs.

## Use Cases

- Refresh application model identifiers: Check provider APIs before updating production model configuration or deployment settings.
- Choose a task-appropriate model tier: Compare fast, default, and heavy recommendations for latency, cost, and capability needs.
- Monitor newly available models: Detect unclassified provider models and review suggested tiers before adding them to the registry.

## Prompt Templates

### List cached models

```
Show the cached models for each provider. Include the fetch timestamp and any provider errors.
```

### Refresh one provider

```
Refresh models for [provider]. Summarize available identifiers and note any authentication or connection errors.
```

### Recommend model tiers

```
Review current models for [provider]. Recommend fast, default, and heavy options with concise reasons based on available metadata.
```

### Audit and classify new models

```
Refresh all enabled providers, detect unknown models, and propose tiers. Do not update the registry until I approve each uncertain classification.
```

## Limitations

- Anthropic, OpenAI, and Gemini queries require valid provider API credentials.
- OpenAI discovery currently filters results to model identifiers beginning with gpt-5.2.
- Tier inference depends on naming patterns and can misclassify unfamiliar model families.
- Ollama discovery only lists models available at the configured host.

## Best Practices

- Review cached timestamps before making network requests.
- Inspect provider errors before using returned model identifiers.
- Approve uncertain tier assignments manually before updating the known-model registry.

## Anti Patterns

- Do not hardcode provider model versions without checking current availability.
- Do not place API credentials in prompts, logs, or shared output.
- Do not auto-classify unfamiliar models when no naming pattern matches.

## Security Audit

- Audited at: 2026-07-13T16:24:07.006\+00:00
- Summary: No malicious intent or prompt injection was found, and most alerts are documentation or language-token false positives. Confirmed risks include broad parent .env loading, credentials in curl arguments or URLs, an unvalidated Ollama host, and a predictable temporary path.

## Stats

- Views: 178
- Downloads: 7
- Favorites: 0
- Popularity score: 0