frontend-template: Skill for Codex

Cursor Rule Refactor

Refactor the project's AI instructions without weakening its existing standards. Treat this as a migration of behavior, not a rewrite from memory.

Workflow

1. Inspect the repository before changing anything.
2. Inventory every file under .cursor/rules/ and .cursor/commands/ if present.
3. Run scripts/audit_cursor_rules.py --root . --output docs/ai/rules-audit-inventory.md when Python is available.
4. Read every rule semantically; do not classify from filename alone.
5. Classify each instruction using references/classification-model.md.
6. Produce a migration map before editing files.
7. Detect duplicates, contradictions, stale patterns, and oversized always-on context.
8. Refactor into the target architecture below.
9. Preserve original rules outside the active .cursor/rules directory until the migration has been verified.
10. Validate that every material original requirement has a destination.
11. Produce docs/ai/rules-migration-report.md using references/migration-report-template.md.

Target architecture

Use this structure as the default. Adapt names to the project when existing conventions are better.

.cursor/
  rules/
    core/
      00-delivery-contract.mdc
      10-project-constraints.mdc
    frontend/
      ...file-scoped engineering rules...
    testing/
      ...file-scoped test rules...
    tooling/
      ...tool-specific rules...
  skills/
    ...repeatable workflows...
  agents/
    ...independent reviewers/verifiers...
docs/
  ai/
    rules-audit-inventory.md
    rules-migration-report.md
  specs/
    ...feature specifications...

Do not create empty categories merely to match the example.

Classification rules

Keep as an always-on rule only when all are true

• The instruction is non-negotiable across almost every task.
• It is short enough to justify permanent context cost.
• Violating it would produce consistently unacceptable output.
• It is not a procedure with several steps.
• It does not describe a temporary persona.