Accept Chrome recorder JSON without a trailing extract step

[cheos71]

Summary

• Chrome side-panel Save/Quick Download writes workflows that end on click/input and use targetText. WorkflowDefinitionSchema rejected those files, so run-workflow-no-ai could not load a recording it just produced.
• Drop the load-time “must end with extract” validator (extract remains the right generation convention for run-as-tool, not a recorder requirement).
• Accept targetText as an alias for target_text so semantic replay sees the captured label.

Test plan

• Load a click-ending Chrome-shaped payload (targetText) via WorkflowDefinitionSchema.model_validate
• target_text snake_case still loads
• Extract-ending workflows still load
• python cli.py run-workflow-no-ai path/to/sidepanel-download.json on a recording that was not manually given an extract step

Made with Cursor

---

Summary by cubic

Fixes loading of Chrome recorder JSON exports so run-workflow-no-ai can replay recordings that end on click/input instead of extract. Now accepts targetText as an alias for target_text.

• Drops the load-time validator that required a trailing extract step; extract remains the recommended generation convention for run-as-tool.

^{Written for commit 5fdae8b. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 2 files

_{Re-trigger cubic}