Spec Developer

自动化 Spec 驱动开发流程 (spec-draft, spec-plan, spec-execute)

Security Scan

The skill's declared purpose (automating spec-driven development) matches most of its instructions, but there are inconsistencies and operational risks (missing install/templates, broad repo write/exec authority, unclear safeguards) that you should review before enabling it.

Purpose & Capability

Name/description match the instructions: drafting specs, planning tasks, and executing tasks (modifying code, running tests) are coherent with a spec-developer skill. The resources it needs (repository files, test script) are appropriate for that purpose.

Instruction Scope

Runtime instructions tell the agent to read/write project files (specs/, specs/tasks.md, .agent/skills/... template, CLAUDE.md), modify/implement code, create session tasks, and run ./tools/run_native_tests.sh. Those actions are powerful (code changes, executing test scripts) and the SKILL.md provides no guardrails (dry-run mode, PRs, scoping to specific paths). The template path referenced (.agent/skills/spec-developer/templates/feature-spec.md) is required but the skill has no install step to ensure that template exists.

Install Mechanism

Instruction-only skill with no install spec; nothing will be written to disk by an installer. This is the lowest-risk install mechanism.

Credentials

The skill declares no environment variables or credentials. The instructions also do not request external API keys. However, executing local test scripts can implicitly rely on environment or local secrets (not declared), so you should confirm the test script's behavior.

Persistence & Privilege

always:false (normal) and autonomous invocation allowed (default). Because the skill can modify code and run tests, allowing it to run autonomously increases risk; consider restricting autonomous invocation or requiring explicit user confirmation before any repository modifications.

What to consider before installing

This skill is broadly coherent with its stated purpose but it can modify your repository and run local test scripts. Before installing: (1) verify the referenced template (.agent/skills/spec-developer/templates/feature-spec.md) actually exists or that the SKILL.md will be updated to provide it; (2) review ./tools/run_native_tests.sh to ensure it is safe and won't leak secrets or perform unexpected network calls; (3) prefer a dry-run or require manual review/PRs rather than direct commits—ensure the skill opens changes as draft commits/PRs instead of pushing directly; (4) limit autonomous invocation or require explicit confirmation before spec-execute runs; (5) backup the repo or run the skill in an isolated clone first. These steps will reduce the chance of unintended or destructive changes.

Spec Developer Skills

本技能集提供一组简短命令，用于快速驱动 Spec 开发流程。

🚀 核心指令

0. spec (帮助)

指令: /spec 行为:

1. 输出 Spec 开发模式的完整帮助信息。
2. 列出可用命令及其说明。
3. 显示当前的 specs/ 目录结构概览。

1. spec-draft (起草)

指令: /spec-draft "功能描述" 行为:

1. 分析用户描述，生成 kebab-case 文件名 (如 specs/features/zero-width-sync.md)
2. 读取模板 .agent/skills/spec-developer/templates/feature-spec.md
3. 填充 Goal, User Stories, Technical Design 等章节
4. 写入文件并展示给用户审查

2. spec-plan (规划)

指令: /spec-plan specs/features/xxx.md 行为:

1. 读取目标 Spec 文件
2. 读取主任务清单 specs/tasks.md
3. 将 Spec 中的 "Tasks Breakdown" 章节提取并追加到 specs/tasks.md 的新里程碑中
4. 确保任务编号 (TASK-XXX) 连续且唯一

3. spec-execute (执行)

指令: /spec-execute specs/features/xxx.md 行为:

1. 读取 Spec 文件和 CLAUDE.md (确保遵循技术红线)
2. 循环执行 Spec 中的每个未完成任务：
   • 使用 TaskCreate 创建会话级任务
   • 编写/修改代码 (遵循 Swift 6 并发规则)
   • 编写单元测试
   • 运行 ./tools/run_native_tests.sh fast
   • 通过后，更新 specs/tasks.md 勾选该任务
3. 所有任务完成后，更新 Spec 头部 Status 为 "Implemented"

使用示例

# 1. 起草新功能
/spec-draft "实现 iOS 17 交互式 Widget"

# 2. 注册到任务清单
/spec-plan specs/features/interactive-widget.md

# 3. 开始自动实施
/spec-execute specs/features/interactive-widget.md