fix(workflows): strip the resolved value before matching switch cases

[jawwad-ali]

Problem

SwitchStep.execute matches with:

str_value = str(value) if value is not None else ""

No strip. But the values a switch dispatches on are overwhelmingly captured command output, and ShellStep stores proc.stdout verbatim (shell/__init__.py:67). So run: echo approve resolves to "approve\n", which matches no approve: case key.

Reproduction on current main (bf88c9f)

stdout stored : 'approve\n'
matched_case  : '__default__'      <-- silently wrong
next steps    : ['fallback']

The switch silently falls through to default: — or, with no default, completes having dispatched nothing — while still reporting COMPLETED.

A workflow author cannot work around it: the registered filters are default / join / map / contains / from_json. There is no trim.

This is already treated as a bug everywhere else

spec-kit strips before matching a resolved string against declared literals in every comparable site:

Site	Code
evaluate_condition	result.strip().lower() — fixed for this exact shell-newline reason
InitStep._resolve_bool	resolved.strip().lower()
workflows/catalog.py	raw_install.strip().lower() in (...)

Switch case keys are such declared literals, and this was the only site not stripping.

Fix

One line, plus the class docstring which said "exact match, string-coerced" and is now accurate.

Narrow. expression_value still reports the raw value, so nothing downstream loses information — pinned by an assertion. A genuine mismatch still falls through: a second test checks "approve-later\n" still routes to default.

The only inputs whose behaviour changes are ones that previously took the wrong branch.

Verification

• Fail-before / pass-after: 2 new-vs-baseline failures with the source reverted; all pass with the fix. Parametrized over "approve\n", "approve\r\n", " approve " and "approve".
• Tests go into the existing TestSwitchStep class — several open PRs touch this file, and a duplicate class name silently shadows.
• Scoped regression: no new failures vs a clean-main baseline captured on bf88c9f9 (21 pre-existing in scope).
• uvx ruff@0.15.0 check src tests → clean

One note on the run: an unrelated test_resume_rejects_malformed_run_state_origin_fields case failed once during verification. I checked rather than assumed — it fails on a different parametrization each time and passes 4/4 in isolation both with and without this change, so it is a pre-existing Windows flake (os.replace PermissionError). A clean re-run of the full gate passed.

---

Written with assistance from Claude Code. Bug found, reproduced, and verified by me on current main.