Mostly lean and command-driven with no concept-explanation fluff, but it restates the spec-merge/校验 steps in prose after the Step 2 numbered list and repeats the directory-move in 退出条件, so not every token earns its place; not a 3 because it could be tightened.

Provides fully executable commands (node "$COMET_STATE" check <name> archive, node "$COMET_ARCHIVE" "<change-name>", git add -A / git commit) and concrete single-choice options that are copy-paste ready.

Clear 0→0b→1→2→3→4 sequence with explicit validation (entry check, exit conditions, non-zero-exit stop) and a feedback loop (archive-reopen back to verify) around an irreversible operation.

Keeps the body an overview and defers detail to one-level-deep, clearly signaled references (comet/reference/scripts.md, decision-point.md, context-recovery.md) rather than inlining it.

Names multiple concrete actions — '确认归档', '合并 delta spec', '恢复 archive 阶段' — rather than vague language, matching the 'lists multiple specific concrete actions' anchor.

Opens with an explicit 'Use when ...' trigger and names the operations the skill performs, so both 'when' and 'what' are answered; the explicit trigger clears the completeness-2 cap.

Natural trigger terms for the niche ('验证已通过', '归档', '恢复 archive 阶段') give good coverage of the conditions a Comet user would actually express; not level 2 because no common trigger phrasing is missing.

Scoped tightly to the Comet change archive phase with distinct triggers, making it unlikely to fire for unrelated skills.