Agent Skills › bacoco/BMad-Skills › bmad-development-execution

bmad-development-execution

GitHub

根据故事文件、架构和测试计划，端到端实现代码功能并执行测试。确保代码变更可审查、透明且可追溯，通过质量门禁后更新文档并通知编排器进入下一阶段。

.claude/skills/bmad-development-execution/SKILL.md bacoco/BMad-Skills

Trigger Scenarios

用户要求实现特定故事或功能用户发出开始编码或开发的指令故事文件已就绪且需进行代码实现

Install

npx skills add bacoco/BMad-Skills --skill bmad-development-execution -g -y

More Options

Non-standard path

npx skills add https://github.com/bacoco/BMad-Skills/tree/main/.claude/skills/bmad-development-execution -g -y

Use without installing

npx skills use bacoco/BMad-Skills@bmad-development-execution

指定 Agent (Claude Code)

npx skills add bacoco/BMad-Skills --skill bmad-development-execution -a claude-code -g -y

安装 repo 全部 skill

npx skills add bacoco/BMad-Skills --all -g -y

预览 repo 内 skill

npx skills add bacoco/BMad-Skills --list

SKILL.md

Frontmatter

{
    "name": "bmad-development-execution",
    "metadata": {
        "outputs": [
            "source-code",
            "test-files",
            "implementation-notes"
        ],
        "triggers": {
            "keywords": [
                "implement",
                "code",
                "develop",
                "build",
                "program",
                "coding",
                "implementation"
            ],
            "patterns": [
                "implement story",
                "develop this",
                "let's code",
                "write the code",
                "start coding"
            ]
        },
        "auto-invoke": true,
        "capabilities": [
            "code-implementation",
            "test-writing",
            "debugging",
            "code-review"
        ],
        "prerequisites": [
            "user-stories",
            "architecture-decisions",
            "test-strategy-doc"
        ]
    },
    "description": "Implements stories with code and tests.",
    "allowed-tools": [
        "Read",
        "Write",
        "Grep",
        "Bash"
    ]
}

Development Execution Skill

When to Invoke

Automatically activate when user:

Says "Implement story X", "Start coding", "Develop this"
Asks "Write the code", "Let's code", "Build [feature]"
Mentions "implement", "code", "develop"
Story file ready (Phase 4)
Uses words like: implement, code, develop, build, program, coding

Specific trigger phrases:

"Implement story [X]"
"Start coding [feature]"
"Develop this story"
"Let's code"
"Write the implementation"
"Build [feature]"

Prerequisites:

Story file exists (from bmad-story-planning)
Architecture patterns defined

Do NOT invoke when:

No story file (use bmad-story-planning first)
Planning needed (use bmad-product-planning first)
Architecture not defined (use bmad-architecture-design first)

Mission

Implement approved stories end-to-end, maintaining transparency, testing discipline, and traceability back to requirements and architecture decisions.

Inputs Required

story: developer-ready story file from delivery-planning skill
architecture_refs: relevant sections of ARCHITECTURE.md or component notes
ux_guidance: UX specs or validations tied to the story
quality_plan: scenarios or gates supplied by quality-assurance skill

Outputs

Code diffs and test results captured in the working repository
Updated story file (Dev Agent Record, status transitions, learnings)
implementation-notes.md - Development summary (generated from assets/implementation-notes-template.md.template)
Summary of changes, tests, and outstanding risks for stakeholders

Template location: .claude/skills/bmad-development-execution/assets/implementation-notes-template.md.template

Process

Confirm prerequisites via CHECKLIST.md and restate story scope.
Plan implementation steps, identifying affected files and tests.
Apply small, reviewable code changes with explanations and references.
Execute required test suites and capture command output verbatim.
Update story documentation, including Dev Agent Record and status.
Summarize work, highlight follow-ups, and notify orchestrator for next steps.

Quality Gates

All items in CHECKLIST.md must pass before code is considered complete. Never mark a story done without full testing evidence.

Error Handling

If prerequisites or environments are missing:

Halt implementation, document the specific blocker, and notify orchestrator.
Provide recommended remediation steps (e.g., refresh artifacts, fix failing baseline tests).
Avoid speculative changes; keep diffs scoped to the approved story.