- Home
- Skills
- Agents & Orchestration
- Model Migrate
Works with the AI Agents you already use
Model Migrate
Find, audit, and migrate deprecated AI models by checking live provider docs and benchmarking output drift.
Free
Model Migrate
Example session with this skill installed
Inventory the models used in this repo and check if any agent models are approaching their shutdown dates based on the latest docs.
- Read your context and instructions
- Compiled the model migrate
Illustrative example with placeholder model names. Real runs use each provider's current deprecation and pricing pages.
Summary
Found 3 model IDs in 5 files. One runtime model shuts down soon, and its listed replacement needs a code change. Nothing has been edited yet.
Model status
| Model | Used in | Kind | Shutdown | Listed replacement | State |
|---|---|---|---|---|---|
provider-model-a | src/grade.ts:14 | runtime, env override | in 42 days | provider-model-b | Urgent |
provider-model-a | src/cost.ts:31 | cost branch | in 42 days | provider-model-b | Urgent |
provider-speech-a | scripts/audio.mjs:8 | script | none listed | none | No date found |
provider-model-c | tests/grade.test.ts:5 | test | not on the page | unknown | No date found |
"No date found" does not mean the model is safe.
Replacement check for provider-model-b
- Endpoint: supported.
temperature: not accepted, so the call must drop it.- Reasons by default: yes. Reasoning tokens count against your output limit, so the current limit may truncate responses.
src/cost.tshas no price for this model, so spend would be reported as 0.
Proposed comparison
6 synthetic cases (weak, typical, strong), 2 runs each, on 2 models: 24 paid calls to one provider. Nothing runs until you approve it.
Next steps
- Approve the 24 calls, or tell me to skip the comparison.
- Confirm which model production really uses.
- After you pick a model, I will update the IDs, the cost branch and the parameters, then run your tests.
Connects securely to your tools. The creator never sees your data.
About this skill
Finding and swapping AI model IDs is simple until a retirement date hits or a replacement model breaks your JSON schemas and doubles your latency. This skill manages the entire migration lifecycle for OpenAI, Anthropic, and Gemini models within your codebase. It moves beyond simple search-and-replace by validating endpoint compatibility, verifying parameter support, and checking live deprecation schedules directly from provider documentation.
What it does
- Codebase inventory: Scans project files to categorize model usage into runtime calls, scripts, test suites, and environment overrides.
- Deprecation tracking: Fetches live shutdown dates and recommended replacements from official OpenAI, Anthropic, and Google documentation.
- Compatibility auditing: Checks if replacement models support existing parameters like
temperatureorreasoningand verifies endpoint availability. - Benchmarking: Runs comparative tests using your project's specific prompts and schemas to detect output drift, latency changes, and cost fluctuations.
- Safe migration: Updates hard-coded IDs, cost logic branches, and feature flags while adjusting parameters for new model architectures.
How it works
- Inventory: Use the grep command to locate all model strings and map their implementation context across the repository.
- Check Dates: Query provider status pages to build a migration timeline and identify models at risk of shutdown.
- Audit: Compare current implementation requirements against the technical specifications and pricing of the proposed replacement.
- Compare: Execute a controlled test script using synthetic inputs to measure performance delta between models.
- Apply: Execute the code changes across the codebase and verify the build.
Frameworks & tools
Works with any codebase using OpenAI, Anthropic, or Gemini APIs. Uses standard Unix utilities like grep for discovery and requires access to provider documentation URLs.
Why this beats prompting it yourself
Generic prompts rely on stale training data for model names and prices. This skill fetches live data from provider docs and uses your actual project prompts to verify that a "cheaper" model won't break your production parsers.
Use cases
- Upgrading from deprecated OpenAI snapshots to the latest stable versions.
- Auditing a legacy codebase to find hidden LLM dependencies before a provider shutdown.
- Benchmarking a move from high-cost models to reasoning-heavy or cost-optimized alternatives.
- Refactoring application logic when switching providers (e.g., OpenAI to Anthropic).
Known limitations
Cannot detect model IDs assembled dynamically at runtime or managed solely in cloud provider dashboards. Benchmarking requires user-provided API credentials and incurs provider costs.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Cannot detect dynamically assembled IDs or dashboard-only settings. Requires manual execution of test scripts for security. Benchmark calls incur provider costs.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 8 days ago
- Free with an Agensi account
Permissions
Terminal: runs grep to find model IDs, and tests after edits. Read and write files: reads project code, skips real .env files, and edits only after approval. Network: reads public provider deprecation, model and pricing pages.