Back to catalog / Agent workflows

Migrate legacy code in an isolated sandbox

Apply explicit migrations to small repositories: test baselines, edit and verify in isolation, then return reviewable patches and an audit summary.

Goal

Apply explicit migrations to small repositories: test baselines, edit and verify in isolation, then return reviewable patches and an audit summary.

Prepare a repository and MIGRATION.md for each task; run baseline tests first. [src-openai-migrate-legacy-code-in-isolated-sandbox]
Stage only that repository and brief into a fresh sandbox for agent edits and checks. [src-openai-migrate-legacy-code-in-isolated-sandbox]
Collect typed results, patches, and audit logs on the host, validate them, then delete the sandbox. [src-openai-migrate-legacy-code-in-isolated-sandbox]

Confirm baseline ran before edits, each task has a result/patch/audit log, and post-migration checks pass.

If a patch or log is missing, inspect host artifact collection and sandbox teardown; do not use final prose as success evidence.

Python 3.10+, Docker or sandbox provider, OpenAI API key, isolated repository fixtures, and migration instructions.

Model keys stay on the host; the sandbox runs commands and edits staged files, while the host writes patch bundles; nothing is auto-applied to a real repo.

The sample uses small fixtures and does not prove a large migration safe; review every patch and build result.

Repository: openai/openai-cookbook
File path: examples/agents_sdk/sandboxed-code-migration/sandboxed_code_migration_agent.ipynb
Pinned commit: 5bd6cb40b6b84b69b2032d810610aa85064102df
Source checked: 2026-10-14 · Reviewed against the pinned upstream source

These values describe the upstream repository, not popularity or likes for an individual skill or playbook.

No metrics snapshot is available yet.

License: MIT