Agent Skills › UniClipboard/UniClipboard › review-strict

review-strict

GitHub

对当前分支或工作区变更进行严格、基于证据的代码审查，不修改文件。重点评估正确性、安全、架构及测试覆盖，提供具体行号引用和修复建议，最终给出合并评估结论。

.agents/skills/review-strict/SKILL.md UniClipboard/UniClipboard

Trigger Scenarios

用户要求进行严格审查用户要求深度审查用户要求高级别审查用户要求进行合并前评估显式调用 $review-strict

Install

npx skills add UniClipboard/UniClipboard --skill review-strict -g -y

More Options

Non-standard path

npx skills add https://github.com/UniClipboard/UniClipboard/tree/v1/.agents/skills/review-strict -g -y

Use without installing

npx skills use UniClipboard/UniClipboard@review-strict

指定 Agent (Claude Code)

npx skills add UniClipboard/UniClipboard --skill review-strict -a claude-code -g -y

安装 repo 全部 skill

npx skills add UniClipboard/UniClipboard --all -g -y

预览 repo 内 skill

npx skills add UniClipboard/UniClipboard --list

SKILL.md

Frontmatter

{
    "name": "review-strict",
    "description": "Perform a strict, evidence-based review of the current branch or working-tree changes without modifying files. Use when the user asks for a strict review, deep review, senior review, merge assessment, or explicitly invokes `$review-strict`."
}

Strict Code Review

Review the current diff first. Inspect unchanged code only when it is necessary to establish the behavior of a changed path.

Prioritize:

Correctness, regressions, state transitions, edge cases, concurrency, cleanup, compatibility, and performance.
Security and privacy, especially plaintext persistence, sensitive logs, excessive permissions, clipboard data exposure, authentication reuse, and raw paths.
Repository architecture, ownership, existing reusable modules, hard-coded values, dead code, and parallel old/new logic.
Cross-platform behavior on macOS, Windows, Linux, iOS, Android, and HarmonyOS where the changed path applies.
Missing or ineffective tests, including duplicated hand-written mocks where an established test helper exists.

For every finding:

Cite the file and exact line.
State the concrete trigger as input or state -> incorrect behavior.
Explain why it matters and the smallest sound correction.
Try to disprove the finding before retaining it.

Order findings by severity: blocking, important, optional. Do not report style preferences or speculative issues. If no real issue remains, say so clearly and note any residual verification gap.

End with one merge assessment: ready to merge, merge after fixes, redesign required, or insufficient evidence. Do not edit, commit, or push.