evalview check --watch (re-run on file change)

[hidai25]

Why

The inner loop today is: edit agent code → save → switch to terminal → run evalview check → read output → back to editor. A --watch flag that re-runs automatically on file change tightens that loop dramatically, and it's the kind of feature people share screenshots of.

What

Add --watch to evalview check that watches a configurable path (default: current directory) and re-runs the check when source files change.

Acceptance criteria

• evalview check --watch runs once, then re-runs on file change
• Configurable include/exclude globs (default: watch *.py / *.yaml under cwd, exclude .evalview/, .venv/, __pycache__)
• Debounce (e.g. 300ms) so a single save doesn't trigger two runs
• Clear console output between runs (clear screen + timestamp header)
• Ctrl+C exits cleanly
• Add watchdog to optional dependencies (pip install evalview[watch])

Hints

• evalview/commands/check_cmd.py is the command file.
• watchdog is the standard Python file-watcher.
• Don't reinvent: wrap the existing check logic in a loop, don't refactor it.

Size

~1-2 hours. Good first issue.