- Home
- Skills
- Agents & Orchestration
- 🔁 Model Migration Checker
More screenshots
Works with the AI Agents you already use
🔁 Model Migration Checker
Review model-version coupling in code, or compare reviewed deployments against exact, dated provider/platform/endpoint migration rules.
$15
🔁 Model Migration Checker
Example session with this skill installed
Save the complete Model Migration Checker v1.1.2 package, run its included risky-sample fixture, and explain what needs review without assuming its parameters are deprecated.
- Read your context and instructions
- Compiled the model migration checker
Packaged fixture run, not customer evidence
Command: python scripts/scan_model_migration.py fixtures/risky-sample
The run returns 11 findings. All locations below are in fixtures/risky-sample/client.py.
- MMC001, medium, lines 1, 2, 3 and 25: four hardcoded model-name/version findings.
- MMC003, medium, line 4: hardcoded token/context-window limit assumption.
- MMC006, low, lines 5 and 27: two per-token cost constant findings.
- MMC002, medium, line 14: max_tokens parameter coupling, REVIEW_CANDIDATE.
- MMC002, medium, line 15: functions parameter coupling, REVIEW_CANDIDATE.
- MMC005, low, line 15: tool/function-schema format coupling.
- MMC004, low, line 28: response parsing tied to one output shape.
Interpretation
Deprecation is not confirmed without provider/platform/endpoint evidence. Coverage is incomplete; environment files are excluded. Review relevant call sites and source documentation before changing them. The separate scoped-rule mode can evaluate supplied dated rules against reviewed deployments, but this fixture run does not execute that mode or test a replacement model.
Connects securely to your tools. The creator never sees your data.
About this skill
Two local review modes
Model Migration Checker package v1.1.2 includes a heuristic scanner and a separate opt-in scoped-rule evaluator. The scanner flags model strings, parameter coupling, token-limit assumptions, response parsing, tool schemas and cost constants. Bare parameter matches are review candidates, not confirmed deprecation.
The evaluator compares reviewed deployment records against dated, source-linked rules you supply. Results distinguish APPLIES, DOES_NOT_APPLY and UNKNOWN_SCOPE. It checks scope and freshness but does not authenticate sources or bundle a universal current deprecation database.
Before you buy
Agent-loaded skill, not a hosted app. Requires Python 3.10+ using the standard library and an agent host that can retrieve and save supporting files and run Python. Agensi supplies access, not execution. Preserve the complete scripts and references folder structure; run the saved Python file, not its source through standard input. A separately obtained download can be extracted with the same structure. Supply reviewed non-secret input. The scanner does not modify source files; the scoped evaluator writes a report to a new local output folder. Neither mode calls models, rewrites source code or proves migration safety.
First run
python scripts/scan_model_migration.py fixtures/risky-sample`Run from the saved package folder, or invoke the saved script by absolute path. Review the sample, then substitute your approved source folder. For the separate scoped mode, read the included scoped-rules reference. Missing, linked or unreadable scanner input produces CANNOT_ASSESS; this link-refusal claim does not cover the separate evaluator.
What changed in v1.1.2
Selected Fable/Mythos identifier patterns, including the literal claude-mythos-preview, and an Anthropic conversation-prefix review checklist. A detected name does not establish model availability, and the checklist does not validate provider signatures. Regression evaluation remains separate and requires actual observations.
Heuristic matches are review candidates, not confirmed deprecation. Save the complete package with scripts/references intact: standard-input execution is unsupported and can fail with a Python error in v1.1.2. Scoped mode needs supplied dated rules and reviewed identities; it does not authenticate sources. Scanner link refusal does not apply to the separate evaluator. Dynamic configuration and runtime behavior remain gaps. No model calls, provider-signature validation or migration-safety certification.
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 today
Requires
Permissions
File Scopes
Reads reviewed non-secret input and editable heuristic patterns. Pattern matches are review candidates, not confirmed deprecation. The optional scoped evaluator uses owner-supplied dated rules and writes reports only to a new local output folder. The scanner does not modify source. No model calls, source rewrites or environment-variable access. This declaration does not grant permission to overwrite existing files.