JustHandled Labs
Menu
// agent-native

Model Migration Checker

Review model-version coupling in code, or compare reviewed deployments against exact, dated provider, platform and endpoint migration rules.

We earn a commission on eligible purchases after a new Agensi signup through these links. We also sell our own skills on Agensi.

Prefer Gumroad? Get Model Migration Checker for $15 USD on Gumroad. Review checkout for tax or currency conversion.

Access: Agent-loaded on Agensi, not a standalone app. Downloadable access applies only where a separate download offer is linked. Agensi package v1.1.2; separate Gumroad download v1.1.1.

Package requirements: Python 3.10 or newer; standard library only. A file-capable agent or terminal that can save the complete package, including scripts and references, and run the saved Python file. Standard-input execution is unsupported. The separate scoped evaluator writes a new local report folder.

You get: Heuristic scanner, editable patterns, separate scoped-rule evaluator, fixtures, tests and rule-contract guidance.

Licence and purchase details

Commercial JustHandled Labs skill. Use for personal or internal team workflows. Do not resell, redistribute, or publish the package as-is.

An AI-agent subscription or setup service is not included.

Agensi access: On November 17, 2026, Claude retrieved all 17 package files through the Agensi creator connection and ran the saved package in local Linux Python: 25 unittest cases passed, one Windows-only junction case was skipped, and standalone checks passed. Fourteen file hashes matched exactly; three differed only in line endings. Creator-route testing, not paid-buyer checkout or Agensi-hosted execution. Save the complete package before running its Python files; the generic wrapper standard-input invocation is unsupported. Scanner link refusal does not extend to the separate scoped evaluator. Other hosts and live provider calls were not tested.

See tested hosts, versions and workflow limits.

Try the included example first

For the separately obtained downloadable package, extract the entire ZIP, keep scripts and references together, and open a terminal in the extracted skill folder. On Windows, use py -3 instead of python if needed.

python scripts/scan_model_migration.py fixtures/risky-sample

What you should see

The packaged risky fixture returns eleven findings for model IDs, coupling and review candidates. The max_tokens and functions matches are REVIEW_CANDIDATE findings, not confirmed deprecations. Exact endpoint and provider evidence are needed for the separate scoped-rule assessment.

Then use your own evidence

Review findings before changing code. For the separate scoped mode, read references/scoped-rules.md and use scripts/evaluate_scoped_rules.py with your reviewed deployment and dated source rules. The package does not fetch current provider rules.

Reviewed non-secret local source for the heuristic scan; supplied deployment and dated rule records for the separate scoped evaluator. No provider calls or source authentication.

Not an automatic migration, a current provider-rules database, or proof that a parameter is deprecated or a migration is safe. The scanner refuses missing, linked or unreadable source inputs; do not treat those refusals as a clean result. The separate scoped evaluator does not enforce the same link refusal.

// choose with context

Is Model Migration Checker the right skill?

Best fit

A maintainer needs a code-review queue before a model change, or has sourced migration rules to evaluate against a reviewed deployment inventory.

It returns

Heuristic code findings, or a separate scoped rule report with APPLIES, DOES_NOT_APPLY and UNKNOWN_SCOPE decisions.

Do not use it as

Not an automatic migration, a current provider-rules database, or proof that a parameter is deprecated or a migration is safe.

Compatibility, access, version, and licence

Reads or accesses
Reviewed non-secret local source for the heuristic scan; supplied deployment and dated rule records for the separate scoped evaluator. No provider calls or source authentication.
Compatibility
Python 3.10 or newer; standard library only. A file-capable agent or terminal that can save the complete package, including scripts and references, and run the saved Python file. Standard-input execution is unsupported. The separate scoped evaluator writes a new local report folder.
Version
Agensi: package v1.1.2 (platform release v1.4), creator-route delivery checked November 17, 2026. Gumroad download: v1.1.1, last checked October 22, 2026.
Licence
Commercial JustHandled Labs skill. Use for personal or internal team workflows. Do not resell, redistribute, or publish the package as-is. Source: Commercial Terms in the released SKILL.md.

What happens next

Open the Agensi listing, check the current offer and connect a supported agent. Agensi loads the skill through its connector; it no longer provides a download. Approve any purchase before use. Check Agensi’s current connection guide before changing an existing connection.

Questions

Does this stay current automatically?

No. Heuristic patterns are editable, and scoped rules need dated sources and exact provider, platform and endpoint scope. Keep them current yourself; a matched parameter alone does not prove deprecation.

Review the evidence before changing models.

Choose your access route above and start with a reviewed example.

Get the Model Migration Checker on Agensi