fix(workflows): handle an unreadable run state in workflow status - #3999

Problem

workflow status <run_id> and workflow resume <run_id> both load the same run state via RunState.load(). A prior fix aligned the two on the FileNotFoundError and ValueError boundaries, but resume also handles OSError and status never gained that handler:

# workflow_resume
except FileNotFoundError: ...
except ValueError as exc: ...
except OSError as exc:          # <-- present here
    err.print(f"[red]Resume failed:[/red] ...")
    raise typer.Exit(1)

# workflow_status
except FileNotFoundError: ...
except ValueError as exc: ...
                                # <-- missing

So an unreadable state.json — wrong permissions, an I/O error, or a directory sitting where the file belongs — escapes workflow status as a raw traceback, while resume on the very same run exits cleanly.

state_path.exists() returns True for a directory, so the existing guard passes and open() raises.

Reproduction

With a run directory whose state.json is a directory, on unmodified main:

workflow status abc123     exit=1  LEAKED PermissionError   output: ''
workflow resume abc123     exit=1  clean                    output: 'Resume failed: [Errno 13] Permission denied: ...'

status prints nothing at all and dumps a traceback; resume prints a clean error. Reproduced through the real Typer CLI, no mocking.

Fix

Add the missing except OSError beside its siblings, using the same _escape_markup + typer.Exit(1) shape as the neighbouring handlers, and routing through err so the message goes to stderr under --json and the stdout JSON stream stays parseable.

Tests

Two regression tests in TestWorkflowCliAlignment, next to the existing status/resume alignment tests:

• test_status_unreadable_run_state_exits_cleanly — end-to-end CLI, a directory in place of state.json
• test_status_json_unreadable_run_state_error_goes_to_stderr — --json stderr routing, mirroring the sibling ValueError test

Both fail without the source change (verified by reverting _commands.py alone) and pass with it.

tests/test_workflows.py: 898 passed, up from 896 on the same tree, with an unchanged set of 20 pre-existing failures — all Windows symlink tests that need elevation and fail identically on unmodified main.

🤖 Generated with Claude Code