AI Model Retirement Fixer (OpenAI, Claude, Gemini, xAI)
A hardcoded `gpt-4o`, a `claude-sonnet-4-5-20250929`, or a `gemini-2.0-flash` keeps working until the provider turns it off.
$5
Works with the AI Agents you already use
AI Model Retirement Fixer (OpenAI, Claude, Gemini, xAI)
Example session with this skill installed
Scan the current directory for deprecated OpenAI and Anthropic models and fix any errors related to reasoning model parameters.
- Read your context and instructions
- Compiled the ai model retirement
I have scanned the codebase and identified two critical issues: one model shutdown and one parameter rejection error for a reasoning model.
Model Migration Audit
| File | Line | Current Model | Severity | Action | Recommendation |
|---|---|---|---|---|---|
src/llm/client.ts | 42 | gpt-4-0314 | ERROR | Replace | gpt-4o (Original retired 2024) |
src/agents/researcher.ts | 89 | o1-preview | ERROR | Fix Params | Remove temperature, top_p |
package.json | 12 | @google/generative-ai | INFO | Upgrade | Move to @google/genai |
Migration Fixes
// src/agents/researcher.ts
// Before:
const response = await openai.chat.completions.create({
model: "o1-preview",
messages: [{ role: "user", content: prompt }],
temperature: 1,
top_p: 1
});
// After:
const response = await openai.chat.completions.create({
model: "o1-preview",
messages: [{ role: "user", content: prompt }],
// temperature and top_p removed: not supported by o1-preview reasoning effort
max_completion_tokens: 5000
});
Next steps
- Review the pricing differences between
gpt-4-0314andgpt-4o. - Run
npm install @google/genaito begin the SDK migration. - Execute
npm testto verify the new parameter configurations.
Connects securely to your tools. The creator never sees your data.
About this skill
Model shutdowns break production code quietly. A hardcoded gpt-4o, a claude-sonnet-4-5-20250929, or a gemini-2.0-flash keeps working until the provider turns it off. Then the obvious replacement rejects max_tokens, temperature or stop.
This skill gives your coding agent:
- A zero-dependency Node scanner (Node 18+, nothing to install). It finds model ids in .js/.ts/.tsx/.mjs files, package.json, .env files and config, and reports
file:line, severity, the official replacement and the source URL. - A dated data file (as of 2026-10-21) with 136 model rows across OpenAI, Anthropic, Gemini API and xAI. Each row was transcribed from the provider's official deprecation or migration page and carries
source_url,as_ofand averifiedflag. - Parameter checks for the changes that commonly break a migration:
- OpenAI:
max_tokens→max_completion_tokens; Responses APIinput/max_output_tokens/output_text; sampling params on reasoning-effort GPT-6 models; tools in Chat Completions on GPT-6. - Anthropic:
temperature,top_pandtop_kare rejected on Claude 4.7+. - xAI:
stopand penalty params are rejected on reasoning models;logprobsis ignored. - Google: the deprecated
@google/generative-aiSDK.
- OpenAI:
- A 7-step workflow: scan, confirm runtime-configured models, pick the replacement, fix params, upgrade SDKs carefully, verify, report.
- A refresh helper that re-fetches the official pages and flags drift, plus a documented weekly re-check process.
Tested on fixture repos for each provider: 9 automated tests, all passing.
Honest limits:
- It's a static scanner. Models chosen at runtime (from a database or secrets) need a human or agent to check.
- Vertex AI, Bedrock and Azure schedules aren't covered.
- Some older OpenAI and Gemini image/live/TTS rows aren't transcribed yet.
- One OpenAI rule (
stop/logit_biason reasoning models) is marked unverified and reported only as a warning. - Dates are a snapshot. The scanner warns when its data is more than 14 days old.
- It suggests fixes. Your agent makes the edits; nothing is rewritten automatically.
What's included
SKILL.md, scripts/scan.mjs, scripts/check-sources.mjs, scripts/build-data.mjs, scripts/render-table.mjs, data/retirements-2026-10-21.json (+ .md table), references/param-changes.md, references/sources.md, REFRESH.md, VERSION.md, CHANGELOG.md.
Partner platforms like Vertex AI, Bedrock, and Azure are out of scope. Static scanning may miss models constructed entirely via dynamic string concatenation at runtime.
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 any payment.
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
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 22 days ago