REQUIRED for feature implementation. Orchestrates 7-phase workflow with user approval gates and TDD enforcement.
Enforces a 7-phase development workflow with mandatory approval gates and TDD. Use for all feature implementation and non-trivial bug fixes to ensure systematic, verified development.
/plugin marketplace add edwinhu/workflows/plugin install workflows@edwinhu-pluginsThis skill inherits all available tools. When active, it can use any tool Claude has access to.
_shared/learnings-template.md_shared/testing-rules.mdFirst, activate the dev workflow and sandbox:
mkdir -p /tmp/claude-workflow-$PPID && touch /tmp/claude-workflow-$PPID/dev_mode && echo "✓ Dev workflow activated"
Announce: "I'm using the dev workflow. Starting with Phase 1: brainstorming."
REQUIRED NEXT STEP: After activation, IMMEDIATELY invoke:
Skill(skill="workflows:dev-brainstorm")
Do NOT ask "What would you like to implement?" - the brainstorm phase handles requirements gathering.
Structured 7-phase workflow for features and non-trivial bug fixes using Task agents and modular skills.
<EXTREMELY-IMPORTANT> ## The Iron Law of Feature DevelopmentYou MUST use the sub-skills. This is not negotiable.
Each phase has a dedicated skill:
/dev-brainstorm - Initial questions/dev-explore - Codebase exploration/dev-clarify - Post-exploration questions/dev-design - Architecture with user approval gate/dev-implement - TDD implementation/dev-review - Confidence-scored review/dev-verify - Runtime verificationThis applies even when:
If you catch yourself about to skip a phase, STOP. </EXTREMELY-IMPORTANT>
| Thought | Why It's Wrong | Do Instead |
|---|---|---|
| "I'll just implement this quickly" | Skipping phases leads to bugs | Follow all 7 phases |
| "I know the architecture" | Must explore first | Use /dev-explore |
| "User will approve anything" | Approval is a gate | Ask explicitly in /dev-design |
| "Code looks fine to me" | Self-review misses issues | Use /dev-review |
| "It should work" | "Should" isn't verification | Use /dev-verify |
flowchart TD
Start(["/dev"]) --> Brainstorm["Phase 1: /dev-brainstorm<br/>Questions → draft SPEC.md"]
Brainstorm --> Explore["Phase 2: /dev-explore<br/>Codebase → key files"]
Explore --> Clarify["Phase 3: /dev-clarify<br/>Questions → final SPEC.md"]
Clarify --> Design["Phase 4: /dev-design<br/>Approaches → PLAN.md"]
Design --> Gate{User approves?}
Gate -->|No| Design
Gate -->|Yes| Implement["Phase 5: /dev-implement<br/>Per-task ralph loops + TDD"]
Implement --> Review["Phase 6: /dev-review<br/>Confidence ≥80"]
Review -->|Issues| Implement
Review -->|Approved| Verify["Phase 7: /dev-verify"]
Verify -->|Fail| Implement
Verify -->|Pass| Complete([Done])
Phase responsibilities:
| Phase | Skill | Does | Outputs |
|---|---|---|---|
| 1 | /dev-brainstorm | Initial questions | Draft .claude/SPEC.md |
| 2 | /dev-explore | Explore codebase, spawn agents | Key files list (10-15) |
| 3 | /dev-clarify | Questions based on exploration | Final .claude/SPEC.md |
| 4 | /dev-design | 2-3 approaches, user picks | .claude/PLAN.md |
| 5 | /dev-implement | Per-task ralph loops + TDD | Working code + tests |
| 6 | /dev-review | Confidence-scored review | Approval or issues |
| 7 | /dev-verify | Runtime verification | Fresh evidence |
THIS SEQUENCE IS MANDATORY. Do not skip any phase.
When user reports a bug, main chat MUST NOT fix it directly.
Even if you know exactly what the fix is. Even if it's one line.
User: "bug: [description]"
Main chat response:
"Using `/dev-debug` workflow."
Then IMMEDIATELY invoke /dev-debug.
All Claude docs go in .claude/ folder (add to .gitignore):
project/
├── .claude/ # gitignored
│ ├── SPEC.md # requirements (phases 1, 3)
│ ├── PLAN.md # chosen approach (phase 4)
│ └── LEARNINGS.md # attempt log (phase 5)
└── ...
Session markers are stored in /tmp/claude-workflow-$PPID/ (auto-cleaned on exit).
/dev-brainstorm - Initial questions and requirements/dev-explore - Codebase exploration with key files/dev-clarify - Post-exploration clarification/dev-design - Architecture approaches with user gate/dev-implement - Orchestrates implementation (main chat)/dev-review - Confidence-scored code review/dev-verify - Runtime verification gate/dev-debug - Systematic debugging methodology/dev-ralph-loop - Per-task loop pattern (main chat)/dev-delegate - Task agent templates (main chat)/dev-tdd - TDD protocol (Task agents)/dev-test - Testing tools reference (Task agents)This skill should be used when the user asks to "create an agent", "add an agent", "write a subagent", "agent frontmatter", "when to use description", "agent examples", "agent tools", "agent colors", "autonomous agent", or needs guidance on agent structure, system prompts, triggering conditions, or agent development best practices for Claude Code plugins.
This skill should be used when the user asks to "create a slash command", "add a command", "write a custom command", "define command arguments", "use command frontmatter", "organize commands", "create command with file references", "interactive command", "use AskUserQuestion in command", or needs guidance on slash command structure, YAML frontmatter fields, dynamic arguments, bash execution in commands, user interaction patterns, or command development best practices for Claude Code.
This skill should be used when the user asks to "create a hook", "add a PreToolUse/PostToolUse/Stop hook", "validate tool use", "implement prompt-based hooks", "use ${CLAUDE_PLUGIN_ROOT}", "set up event-driven automation", "block dangerous commands", or mentions hook events (PreToolUse, PostToolUse, Stop, SubagentStop, SessionStart, SessionEnd, UserPromptSubmit, PreCompact, Notification). Provides comprehensive guidance for creating and implementing Claude Code plugin hooks with focus on advanced prompt-based hooks API.