fix(cli): read run-workflow-csv inputs as strings to preserve leading zeros

[salapenar-max]

Problem

run-workflow-csv loads the input CSV with pd.read_csv(csv_path) (default dtype inference), and _execute_single_workflow then does str(raw_value) for a string-typed input. When pandas infers a numeric dtype, string identifiers are corrupted:

• a zip code 01234 → int64 1234 → "1234" (leading zero lost)
• 00089 → "89"
• a mixed column (1002, 1002.0) → float64 → "1002.0"

The corrupted value is then substituted into the workflow's {placeholders}, so forms get filled with the wrong identifier (zip, phone, SKU, account number, …). It fails silently — no error is raised.

Fix

Read the CSV with dtype=str so string inputs keep their exact text. The existing per-field conversion still coerces number/bool inputs (float(...), truthy parse), and empty cells remain NaN, so the pd.isna(...) required-field check is unchanged.

-            df = pd.read_csv(csv_path)
+            df = pd.read_csv(csv_path, dtype=str)

Test

Adds workflows/tests/test_run_workflow_csv_string_dtype.py: string fields keep leading zeros; number/bool fields still convert; an empty required cell is still detected as missing.

Note

Prepared with AI assistance and verified locally against the current default branch.

🤖 Generated with Claude Code

---

Summary by cubic

Fixes run-workflow-csv silently corrupting string-typed inputs like zip codes (01234 → 1234) because pandas inferred numeric dtypes and applied default NA detection. Reads the CSV with dtype=str, keep_default_na=False, and na_values=[''] so string inputs keep exact text; number/bool fields still convert, and empty required cells are still detected as missing.

• Literal cells like NA (country code) or null stay string text for string fields, but count as missing for number/bool fields.
• Input-validation failures (empty required cells, conversion errors) now carry a failure_type and show in the run summary as form validation with "check CSV data" guidance, instead of KeyErroring or printing as generic failures.
• Extracts loading and per-field conversion into workflow_use/csv_inputs.py for unit-testability; adds test_run_workflow_csv_string_dtype.py with a __main__ runner, registered in tests/run_all_tests.py.

^{Written for commit 1ceb716. Summary will update on new commits.}

[cubic-dev-ai]

All reported issues were addressed across 2 files

<sub>Reply with feedback, questions, or to request a fix.

Fix all with cubic | Re-trigger cubic</sub>

[cubic-dev-ai]

All reported issues were addressed across 4 files (changes from recent commits).

<sub>Reply with feedback, questions, or to request a fix.

Fix all with cubic | Re-trigger cubic</sub>

[cubic-dev-ai]

All reported issues were addressed across 3 files (changes from recent commits).

<sub>Tip: Review your code locally with the cubic CLI to iterate faster.

Fix all with cubic | Re-trigger cubic</sub>