using-meld

MELD (Methodology for Engineering Lifecycle & Development) provides a complete methodology for structured product development — from planning through execution to review. It includes:

• Complexity routing — Match approach depth to problem complexity
• Spec engineering — Structured Given/When/Then specifications with ready-for-dev standards
• Adversarial review — Information-asymmetric code review using subagents
• Execution skills — TDD, debugging, verification, parallel agents, worktrees

When to Use MELD Skills

| Situation | MELD Skill | Pairs With | |-----------|-----------|------------| | User describes a feature or asks "build X" | meld:meld-complexity-assessment | — | | Feature needs a spec before coding | meld:meld-quick-spec | — | | Ready to implement (with or without spec) | meld:meld-quick-dev | TDD built-in (meld:meld-tdd) | | Writing acceptance criteria | meld:meld-spec-engineering | — | | Bug needs investigation | meld:meld-debugging | meld:meld-tdd | | Claiming work is complete | meld:meld-verification | — | | 2+ independent tasks | meld:meld-parallel-agents | — | | Starting isolated work on a ticket | meld:meld-worktrees | Built into quick-dev | | Code needs cleanup after implementation | meld:meld-code-simplifier | meld:meld-verification | | Implementation complete, need review | meld:meld-adversarial-review | meld:meld-verification | | Implementation complete, capture learnings | meld:meld-retrospective | Built into quick-dev | | Need an output template (tech-spec, PRD, etc.) | meld:meld-artifact-templates | — |

When NOT to Use MELD Skills

• Simple bug fixes with obvious cause — Go straight to TDD. If cause is NOT obvious, use meld:meld-debugging
• One-line changes — Just do it
• Exploration/research — Use exploration tools directly
• Refactoring with no behavior change — Execution skills suffice

Available Skills

Flow Skills (invoke these to run a workflow)

• meld-quick-spec — Conversational spec engineering: understand → investigate → generate → review. Produces a ready-for-dev tech spec.
• meld-quick-dev — Implementation flow: setup & mode detection → context gathering → per-task subagent loop (TDD + simplify + spec review + code review) → final verification → completion & retrospective.
• meld-complexity-assessment — Evaluate complexity signals and route to the right depth of planning.

Methodology Skills (invoked by flow skills or standalone)

• meld-tdd — Test-driven development: Iron Law, Red-Green-Refactor cycle, rationalizations, red flags. Built into quick-dev's execution loop.
• meld-debugging — Systematic debugging: 4-phase root cause methodology (investigate → analyze → hypothesize → implement). Activated by quick-dev halt conditions. Includes 3 supporting docs (root-cause-tracing, defense-in-depth, condition-based-waiting).
• meld-code-simplifier — Code simplification pass: subagent reviews modified code for clarity, consistency, and maintainability. Built into quick-dev's per-task loop.
• meld-verification — Verification before completion: 5-step gate function requiring fresh evidence for every completion claim. Invoked at quick-dev Phase 4.
• meld-parallel-agents — Parallel agent dispatch: identify independent domains, craft focused prompts, dispatch via Task tool, review and integrate. Optional in quick-dev Phase 3.
• meld-worktrees — Git worktree creation with ticket-based branch naming. Auto-creates worktrees for beads tickets. Invoked at quick-dev Phases 1 and 5 (completion).
• meld-spec-engineering — Given/When/Then acceptance criteria format and ready-for-dev standards.
• meld-adversarial-review — Code review with information asymmetry using subagents.
• meld-retrospective — Post-implementation retrospective: categorizes findings patterns, measures spec accuracy, captures estimation signals, persists learnings to project memory. Built into quick-dev Phase 5.

Reference Skills

• meld-artifact-templates — Index of 8 output templates (tech-spec, story, PRD, product brief, architecture decision, epics, UX design, sprint status).

Slash Commands

• /quick-spec [ticket-id] — Start the quick-spec flow (optionally against a beads ticket)
• /quick-dev [ticket-id] — Start the quick-dev flow (optionally against a beads ticket)
• /assess-complexity — Run complexity assessment

Self-Contained Methodology

MELD owns all its methodology and execution skills — from planning (spec engineering, complexity routing) through execution (TDD, debugging, verification, parallel agents, worktrees) to review (adversarial review). No external plugin dependencies are required.