Remove a dead case pattern in common.sh (SC2221/SC2222)

[jzayas03]

Description

resolve_template_content() in scripts/bash/common.sh rejects unsafe manifest paths with:

/*|*../*|../*) manifest_file="" ;;

The third alternative can never match: *../* already matches every string ../* does. shellcheck reports SC2221 ("this pattern always overrides a later one") and SC2222 ("this pattern never matches") on that line. This PR removes the dead alternative — a one-line change with no behavioural effect.

Verified behaviour-identical by classifying these inputs with the old and new pattern: /etc/passwd, ../x, a/../b, ../, x/../../y, .., a/.., ..a, a..b, ./x, templates/x.md. Every one produces the same reject/keep result before and after. After the change, shellcheck reports zero SC2221/SC2222 findings on the file and adds nothing new.

These are warnings, so lint.yml's --severity=error does not currently see them. This is one of the findings standing between the current threshold and the tightening described in CONTRIBUTING.md. Related: #1049.

Testing

• Tested locally with uv run specify --help
• Ran existing tests with uv sync && uv run pytest — 7883 passed, 195 skipped, 1 failed. The failure is tests/integrations/test_events.py::TestCommandRunner::test_ps_variant_prefixed_with_powershell_launcher, and it fails identically with this change reverted: it asserts a pwsh/powershell -File prefix and pwsh is not installed on the machine I ran this on (macOS). It is pre-existing and environmental, not caused by this PR; your macos-latest runner has pwsh and should be green.
• Tested with a sample project — not done, deliberately. The change is proven behaviour-identical (table above), so running commands through an agent could not observe any difference. Mapping per CONTRIBUTING.md for completeness:

Test selection reasoning

Changed file	Affects	Test	Why
scripts/bash/common.sh	every command invoking create-new-feature.sh, check-prerequisites.sh, setup-plan.sh, setup-tasks.sh, resolve-template.sh (all source common.sh)	none run	the diff is a no-op on behaviour; resolve_template_content() accepts and rejects exactly the same inputs

If maintainers would still like the manual agent runs, say so and I will do them.

AI Disclosure

• I did not use AI assistance for this contribution
• I did use AI assistance (describe below)

This PR was produced by Claude (Anthropic's AI coding agent) operating under my direction: it found the finding while running shellcheck over a project that vendors these scripts, verified the fix, wrote the commit and this description. I reviewed and approved it before it was opened. Any follow-up comments from this account on this PR may also be AI-drafted, and I will keep that disclosed.

🤖 Generated with Claude Code