Sample Codemod

Operational Instructions for sample-codemod

Instructions for AI agents to read, adapt, run, and execute this codemod workflow safely.

Reading the Workflow Graph

Parse workflow.yaml. The execution graph is:

1. apply-transforms — AST transformations via scripts/codemod.ts
2. ai-tricky-cases — depends_on: [apply-transforms]; AI step for closure-dependent cases
3. publish — depends_on: [ai-tricky-cases]; push and create PR when params.publish_pr is true

Run nodes in this topological order. Steps inside each node run sequentially.

Step-by-Step Execution

1. Execute each node in topological order.
2. After each change-producing step: Run validation before proceeding.
   • pnpm test — jssg tests
   • pnpm run check-types — typecheck
   • npx codemod workflow validate workflow.yaml — workflow schema
3. Review diffs. Proceed only when validation passes and changes look correct.
4. If validation fails or diffs are wrong, fix and retry before continuing.

AI Steps

When params.run_ai_step is true, the ai-tricky-cases node runs an AI step. Treat the AI step prompt (in workflow.yaml) as context for edge cases that the AST codemod cannot safely handle—e.g. closure-dependent nested components.

• Apply AI changes only where the workflow indicates (inside ai-tricky-cases).
• Keep AI edits localized and reviewable. Add TODOs if unsure.
• Do not re-describe the migration; the prompt already covers the goal and constraints.

Adapting the Workflow and Scripts

To adapt for user code patterns:

• Include/exclude globs — Adjust include and exclude in the js-ast-grep step to match the target project layout.
• Script logic — Modify scripts/codemod.ts to add or refine patterns. Run tests after changes.
• Params — Use run_ai_step=true when the codebase has closure-dependent nested components that need manual/AI resolution.