Token导航 LogoToken导航TokenDH.com
研究检索需要联网github未标认证来源可访问许可证需确认审计通过

spec-driven-review规范驱动的审查

Agent Skill

spec-driven-review 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

838

周安装

36

GitHub Stars

公开资料未说明

下载量

294
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:spec-driven-review(规范驱动的审查)
来源仓库:https://github.com/kw12121212/auto-spec-driven
仓库路径:skills/spec-driven-review
安装命令:
npx skills add https://github.com/kw12121212/auto-spec-driven --skill spec-driven-review
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/kw12121212/auto-spec-driven --skill spec-driven-review

简介

用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词、任务场景或来源线索快速定位候选结果。

  • 可结合来源仓库、安装命令和原始 README 继续核验具体用法。
  • 安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。
  • 通过 npx skills add 命令从指定 GitHub 仓库安装 spec-driven-review 技能。
  • 当前无额外说明,需参考原始 SKILL.md 获取完整功能细节。

SKILL.md

You are reviewing the code quality of a completed spec-driven change.

This Skill's Commands

If you cannot remember the exact command used by this skill, look it up here before running anything. Do not guess.

modify: node {{SKILL_DIR}}/scripts/spec-driven.js modify
apply: node {{SKILL_DIR}}/scripts/spec-driven.js apply <name>
audit-spec-mapping-coverage: node {{SKILL_DIR}}/scripts/spec-driven.js audit-spec-mapping-coverage <spec-path> [--implementation <repo-path> ...] [--tests <repo-path> ...]
audit-unmapped-spec-evidence: node {{SKILL_DIR}}/scripts/spec-driven.js audit-unmapped-spec-evidence [--implementation <repo-path> ...] [--tests <repo-path> ...]

Prerequisites

The .spec-driven/ directory must exist at the project root. Before proceeding, verify:

ls .spec-driven/

If this fails, the project is not initialized. Run /spec-driven-init first.

Steps

  1. Select the change — run node {{SKILL_DIR}}/scripts/spec-driven.js modify to list active changes. Ask which change to review. If already specified, use it.
  2. Confirm readiness — run: node {{SKILL_DIR}}/scripts/spec-driven.js apply <name> If remaining > 0, stop — the change is not ready for review. Suggest /spec-driven-apply first.
  3. Load context — read:

- .spec-driven/changes/<name>/proposal.md — scope and unchanged behavior - .spec-driven/changes/<name>/specs/ — delta specs describing the intended behavior changes - .spec-driven/changes/<name>/design.md — approach and decisions - .spec-driven/changes/<name>/tasks.md — what was implemented - .spec-driven/changes/<name>/questions.md — resolved answers that may explain decisions or tradeoffs - .spec-driven/config.yaml — project context and rules, including test rules and any fileMatch entries - mapping frontmatter from relevant main and delta spec files

  1. Identify changed files — from the completed tasks and mapping frontmatter, determine which files were created or modified. Read each file fully, including mapped implementation and test files for the relevant spec files.
  2. Identify specialized review checklists — before issuing findings, classify the change using evidence from the proposal, delta specs, design, tasks, questions, changed files, mapping frontmatter, and repository context. Apply every checklist that fits; checklist routing is additive and never replaces the baseline review checks. Treat reviewed code as author-agnostic: do not trust or soften review because the code may have been generated by AI, previously produced by you, or written by a human. Supported checklist types:

- Security-sensitive: changes involving authentication, authorization, permissions, secrets, tokens, input boundaries, data exposure, dependency trust, or safe failure behavior. - UI: changes involving user-visible screens, layout, interaction state, accessibility, keyboard behavior, responsive behavior, or visual regressions. - DX: changes involving commands, prompts, generated artifacts, documentation, operator messages, error messages, setup, or local workflow ergonomics. - Migration: changes involving data transformation, filesystem or schema transitions, backwards compatibility, idempotency, rollback, or partial failure behavior. - API: changes involving public or internal contracts, command arguments, JSON shapes, error semantics, validation behavior, versioning, or caller impact. - Maintenance: changes involving scheduled or manual maintenance, dependency upkeep, generated files, repository hygiene, repeatability, or avoiding unrelated churn.

  1. Review code quality — for each changed file, run the baseline review checks plus any specialized checklist checks identified in Step 5:

- Impartiality: judge the code that exists, not who wrote it; do not suppress, soften, or skip findings for AI-authored or self-authored code - Readability: clear naming, reasonable function length, no unnecessary complexity - Security: no injection vulnerabilities, no hardcoded secrets, proper input validation at system boundaries - Error handling: appropriate error handling for external calls and user input; no swallowed errors - Performance: no obvious N+1 queries, unnecessary allocations, or blocking calls in async contexts - Best practices: follows the project's conventions (from config.yaml context), no dead code, no debug artifacts left behind - Security-sensitive checklist: verify authorization boundaries, secret handling, injection resistance, data exposure, validation, and safe failure behavior. - UI checklist: verify key interactions, accessibility, layout stability, state transitions, responsive behavior, and user-visible regression coverage. - DX checklist: verify command ergonomics, documentation or prompt clarity, actionable errors, setup compatibility, and consistency with the existing workflow. - Migration checklist: verify idempotency, backward compatibility, rollback or recovery expectations, partial failure handling, and data preservation. - API checklist: verify contract compatibility, validation, error semantics, versioning or migration expectations, and caller impact. - Maintenance checklist: verify repeatability, dependency or generated artifact safety, repository hygiene, and absence of unrelated churn.

  1. Check test quality — read the test files associated with this change:

- Do tests cover the key scenarios from the delta specs? - Are tests independent and repeatable? - Do tests follow rules.test from config.yaml? - Are mapping.tests entries current and useful for future verification? - Do tests cover the relevant specialized checklist risks when those risks are observable and testable?

  1. Audit mapping quality — for each touched spec file relevant to this change:

- Compare mapping.implementation and mapping.tests against the change's primary implementation files and directly verifying test files - Use the smallest confident evidence set from changed files, delta specs, completed tasks, and mapped files already read - Run node {{SKILL_DIR}}/scripts/spec-driven.js audit-spec-mapping-coverage <spec-path> [--implementation <repo-path>...] [--tests <repo-path>...] when it helps make the comparison explicit - Run node {{SKILL_DIR}}/scripts/spec-driven.js audit-unmapped-spec-evidence [--implementation <repo-path>...] [--tests <repo-path>...] when it helps confirm whether reviewed implementation or test files are missing from all main-spec mappings - Report stale or misleading mappings as at least SHOULD FIX - Escalate to MUST FIX when the mismatch would materially mislead future verification or archive readiness - If the unmapped audit shows that a primary implementation file or directly verifying test file for this reviewed change is missing from all main-spec mappings, treat that as MUST FIX - If the unmapped audit only finds files outside this review scope or weakly related candidates, report them separately as SHOULD FIX or repository debt

  1. Output a review report: MUST FIX (blocks archive): - [list or "none"] SHOULD FIX (recommended): - [list or "none"] NITS (optional): - [list or "none"]
  2. Recommend next step:

- If MUST FIX issues: address them before archiving - If only SHOULD FIX: ask user if they want to address them or proceed - If clean: suggest /spec-driven-archive <name>

Rules

  • Read every changed file before commenting on it — never review code you haven't read
  • Focus on real issues, not style preferences already handled by linters
  • Review code based on evidence in the code and tests, never on authorship or assumed intent
  • MUST FIX = security vulnerabilities, data loss risks, broken functionality
  • SHOULD FIX = maintainability concerns, missing error handling, unclear logic
  • NITS = naming suggestions, minor simplifications, documentation gaps
  • Do not re-review code that was not changed by this change
  • Respect config.yaml rules — violations of project rules are SHOULD FIX at minimum
  • Report stale or misleading spec mappings when they would make future verification unreliable
  • Specialized checklist findings use the same severity model as the baseline review: MUST FIX for archive blockers, SHOULD FIX for recommended fixes, and NITS for optional cleanup
  • Never give self-authored or AI-authored code lighter treatment than human-authored code

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

36.95%
按下载量换算109

Claude

30.46%
按下载量换算90

Cursor

16.14%
按下载量换算47

Gemini CLI

9.78%
按下载量换算29

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills