fix(workflows): reject a non-string prompt in prompt-step validate()

[Noor-ul-ain001]

Summary

PromptStep.execute str()-coerces config['prompt'] and dispatches the result to the integration CLI as the model's instructions:

prompt = evaluate_expression(prompt_template, context)
if not isinstance(prompt, str):
    prompt = str(prompt)   # <-- a list/None/int becomes its Python repr

But PromptStep.validate only checked that prompt was present, not that it was a string — the exact parity gap the sibling ShellStep already closes for its run field (which cites "the command-step input/options and gate options type checks" as precedent).

So a YAML authoring slip like prompt: [review, this] or a bare prompt: (YAML null) passed validation, then execute sent the Python repr ("['review', 'this']", "None") to the LLM verbatim — silently wrong instructions, no error, and a COMPLETED status. The engine does not auto-validate step config (load_workflow explicitly defers validation to validate_workflow()/engine.validate()), so validation is the only place this surfaces before dispatch.

Fix

Extend validate to reject any non-string prompt, mirroring the shell-step run and command-step input/options type checks:

elif not isinstance(config["prompt"], str):
    errors.append(
        f"Prompt step {config.get('id', '?')!r}: 'prompt' must be a "
        f"string, got {type(config['prompt']).__name__}."
    )

A {{ ... }} expression is still a str, so it stays valid.

Testing

• Adds test_validate_rejects_non_string_prompt (parametrized over null/list/int/dict), mirroring TestShellStep::test_validate_rejects_non_string_run.
• Adds test_validate_accepts_expression_prompt to confirm an expression prompt still validates.
• Full TestPromptStep + validation suite passes (85 selected, 0 failures).

🤖 Generated with Claude Code