Trace configuration-file migrations

Use when a config, workspace, or instruction file disappears after an upgrade and the task is to explain what changed, when, why, and whether local data survived.

Procedure

1. Inspect current local state first.

   • Test whether the old file exists.
   • Locate the replacement section or file.
   • Record modification times for the replacement and any migration backup.
   • Hash the backup when integrity matters.

2. Establish upstream intent.

   • Find the introducing PR or commit.
   • Inspect its changed-file list before opening large diffs.
   • Read the PR description and only the relevant source, template, and documentation files.

3. Establish release timing.

   • Read tag name and publication time from release metadata.
   • Do not rely on a truncated release body as proof.
   • Distinguish merge date, first packaged prerelease, and later migration refinements.

4. Read migration code, not only release notes.

   • Identify the source file, destination, backup/archive path, deletion point, and failure ordering.
   • Check how empty or stock templates differ from customized content.
   • Check whether runtime retains a fallback reader for the legacy file.
   • Check migration scope: workspace root, multiple agents, or nested directories.

5. Compare docs and templates with implementation.

   • Confirm the new canonical location and whether it controls behavior or merely supplies model guidance.
   • Treat source as decisive for edge cases; use docs for the supported contract.

6. Audit stale custom guidance.

   • Search active instruction files for references to the legacy path.
   • Inspect rewrite tables when migration updates known wording.
   • Expect exact-match rewrites to miss custom paraphrases; report those separately from data migration.

7. Report in this order:

   • conclusion: deleted, relocated, merged, or unsupported;
   • timeline: merge, release, later refinements;
   • rationale;
   • migration behavior and recovery path;
   • verified local outcome;
   • remaining stale references or out-of-scope nested files.

Pitfalls

• File absence alone does not prove data loss.
• Release notes alone do not establish migration ordering or rollback behavior.
• A backup alone does not prove destination content was written.
• Successful root migration does not prove nested files were migrated.
• Automated wording rewrites may cover stock text but leave customized guidance contradictory.
• Large API responses can truncate the only relevant evidence; narrow by changed path or metadata.

Verification

Verify all applicable invariants:

• the legacy root file is absent or intentionally retained;
• migrated content appears at the documented destination;
• the backup exists and its hash is stable;
• timestamps align with the migration event;
• current docs name the destination;
• migration source confirms archive/write/delete ordering;
• active instructions contain no unreported legacy references.

If any invariant cannot be checked, label it unknown rather than inferring success.