Generated openspec-archive-change skill archives even when the user selects Cancel

[xfxf]

The v1.4.1 generated accelerators (openspec init --tools ...) include an archive skill (.claude/skills/openspec-archive-change/SKILL.md, .codex/skills/..., .claude/commands/opsx/archive.md) whose step 4 offers a Cancel option when delta specs are already synced:

Prompt options: ... If already synced: "Archive now", "Sync anyway", "Cancel"

but the same step then instructs:

Proceed to archive regardless of choice.

An agent following the skill literally will move changeRoot into openspec/changes/archive/... even after the user selected Cancel. Suggested fix: make Cancel stop before step 5; only the explicit archive/sync choices proceed.

(Found while adopting OpenSpec; we patched our committed copies locally with exactly that wording.)

[clay-good]

Partially fixed already, and PR #1398 closes the rest.

The single-change path you quoted was fixed on main by #1394 (commit b474f81, unreleased as of v1.6.0) — archive-change.ts now routes the answer explicitly:

Route on the answer:
- "Cancel" — stop, do not archive
- "Archive without syncing" or "Archive now" — proceed to archive
- "Sync now" or "Sync anyway" — sync, then verify (below)
- Anything else — ask again rather than archiving

The same hole survived in the bulk variant (openspec-bulk-archive-change / opsx:bulk-archive): step 7 offers "Cancel", step 8 archives every selected change with no route honoring it. #1398 mirrors the wording above into both bulk template bodies and adds a regression test, so this issue closes with that PR.

Thanks for the clear report — the exact wording you suggested is what landed.

[clay-good]

Correction to my note above: I misattributed the single-change fix. It landed in #1357 (46a4d78, 2026-07-17), whose commit body reads "archive-change: choosing 'Cancel' at the sync prompt now stops the flow". #1394 only reworded that into the current Route on the answer: block while fixing the sync race (#1393). Neither has shipped — both are post-v1.6.0.

The bulk half is fixed in #1398, which also adds a regression test covering all four archive templates (bulk skill, bulk command, single skill, single command) — the single-change routing was correct but unpinned, so this issue's original report had no test guarding it until now.