The body is lean: no explanation of what git or hooks are, just the blocked-command list, sequenced steps, and necessary JSON/verify snippets, so every token earns its place per the score-3 anchor.

It provides copy-paste-ready settings.json blocks for both project and global scope, exact target paths, 'chmod +x', and a concrete verification command, meeting the 'fully executable, copy-paste ready' anchor.

Five explicitly sequenced steps culminate in a Verify step with a test command and expected exit code 2 / BLOCKED stderr output, an explicit validation checkpoint; it is not the score-2 case because validation is present and concrete rather than missing or implicit.

The hook implementation is appropriately split into the referenced bundle file [scripts/block-dangerous-git.sh] (confirmed to exist), a one-level-deep, clearly signaled reference, keeping the body an organized overview rather than a monolith.

The description enumerates concrete commands to block ('push、reset --hard、clean、branch -D') and the concrete action of installing a PreToolUse hook, matching the 'lists multiple specific concrete actions' anchor rather than the vaguer score-2 example.

It states both what ('设置 Claude Code hooks...阻止危险 git commands') and when ('适用于用户想...时'), with an explicit trigger clause equivalent to 'Use when', satisfying the both-what-and-when requirement that would otherwise cap at 2.

Natural user phrasing is well covered via '防止破坏性 git 操作', 'git safety hooks', and '阻止 git push/reset', plus the raw command names users would actually say; this exceeds the 'some relevant keywords' bar at score 2.

The niche is narrow and specific — blocking destructive git commands via Claude Code hooks — with triggers unlikely to fire for unrelated skills, matching the 'clear niche with distinct triggers' anchor.