Back to list
ksterx

auto-dev

by ksterx

0🍴 0📅 Jan 22, 2026

SKILL.md


name: auto-dev description: | Autonomous development workflow for code implementation, bug fixes, feature development, and refactoring. Automatically determines task complexity and adjusts workflow accordingly. Use when: implementing features, fixing bugs, refactoring code, or any development task.

Autonomous Development Workflow

You are an autonomous development orchestrator. Analyze tasks, determine complexity, and execute the appropriate workflow with minimal human intervention.

References (read as needed)

ReferenceWhen to Read
references/templates/spec.mdWhen generating spec.md
references/templates/plan.mdWhen generating plan.md
references/templates/tasks.mdWhen generating tasks.md
references/agent-prompts.mdWhen invoking agents
references/handoff-schema.mdWhen passing context between agents
references/gemini-review.mdWhen performing Gemini review
references/stacks/typescript.mdTypeScript/React projects (auto-detect)

Preflight: Stack Detection

Before complexity assessment, detect project stack and read relevant stack reference:

Detection CriteriaStack File
package.json has next in dependencies + tsconfig.jsonreferences/stacks/typescript.md (Next.js mode)
package.json has react (no next) + tsconfig.json or .tsx filesreferences/stacks/typescript.md (React-only mode)

Detection notes:

  • Check dependencies or devDependencies entries (not substring match)
  • Also accept tsconfig.*.json variants
  • Skip if project already has established different choices (e.g., Redux, MUI)

Apply stack conventions throughout implementation, respecting existing project choices.


Phase 0: Complexity Assessment

LevelCriteriaSDDGemini
Trivial1-2 lines, single file, obviousNoNo
SimpleClear, 1-3 files, existing patternsNoNo
StandardNew functionality, multiple files, design neededYesYes
ComplexLarge scope, architectural impact, ambiguousYesYes

Output:

Complexity: [LEVEL]
Reasoning: [Brief justification]

Workflow by Complexity

Trivial

Implement directly → Verify → Complete

Simple

explorer(summary) → implementer → tests → reviewer → auto-fix → Complete

Standard (with SDD)

EXPLORATION
  explorer(detailed) → [Gemini] → validate

SPECIFICATION
  architect(spec) → [Gemini] → 👤 APPROVE
  architect(plan) → [Gemini] → 👤 APPROVE
  architect(tasks) → [Gemini] → 👤 APPROVE

IMPLEMENTATION
  implementer → tests → [Gemini] → validate

REVIEW
  reviewer(spec-compliance) → [Gemini]
  reviewer(standard) → [Gemini] → auto-fix → Complete

Complex (with SDD)

EXPLORATION
  explorer(scoped) x2-3 parallel → [Gemini each]
  👤 REVIEW: clarify requirements

SPECIFICATION
  architect(spec) → [Gemini] → 👤 APPROVE
  architect(plan + approach) x2-3 parallel → [Gemini each] → 👤 CHOOSE & APPROVE
  architect(tasks) → [Gemini] → 👤 APPROVE

IMPLEMENTATION
  implementer(phased) → tests each phase → [Gemini]

REVIEW
  reviewer(spec-compliance) → [Gemini]
  reviewer(scoped) x3 parallel → [Gemini each] → auto-fix → Complete

SDD Documents

specs/{feature-slug}/
├── spec.md    # What (requirements, acceptance criteria)
├── plan.md    # How (architecture, components)
└── tasks.md   # Steps (phased checklist)

Naming: kebab-case, English. Optional: YYYYMMDD-feature-slug


Agent Modes Quick Reference

AgentModes
code-explorerdetailed, summary, scoped:[entry/data/cross-cutting]
code-architectspec, plan, tasks, full, parallel:[minimal/clean/pragmatic]
code-implementer(default)
code-reviewerstandard, spec-compliance, bugs, quality, conventions

For full prompts, read references/agent-prompts.md


Gemini Review (Standard+ only)

After each agent output:

  1. Call mcp__gemini__gemini-brainstorm with summary
  2. Evaluate response neutrally and fairly
  3. Apply valid fixes, ignore subjective/style-only feedback
  4. Present to human with applied/ignored notes

For full protocol, read references/gemini-review.md


Human Review Points (Standard+)

DocumentApproval Criteria
spec.mdRequirements complete, AC clear
plan.mdArchitecture sound, approach approved
tasks.mdTasks atomic, dependencies correct

Iteration Budget

ComplexityBudget
Trivial0
Simple2
Standard6
Complex10

Definition of Done

  • All requirements from spec.md implemented
  • All acceptance criteria verified
  • All tasks in tasks.md checked
  • Lint/type-check pass
  • Tests pass
  • Spec compliance verified

Output Format

## Summary
- **Task**: [Request]
- **Complexity**: [Level]
- **Feature**: [slug]
- **Documents**: specs/{slug}/ ✅
- **Files modified**: [List]
- **Tests**: [Pass/Fail]

Score

Total Score

40/100

Based on repository quality metrics

SKILL.md

SKILL.mdファイルが含まれている

+20
LICENSE

ライセンスが設定されている

0/10
説明文

100文字以上の説明がある

0/10
人気

GitHub Stars 100以上

0/15
最近の活動

3ヶ月以内に更新がある

0/10
フォーク

10回以上フォークされている

0/5
Issue管理

オープンIssueが50未満

+5
言語

プログラミング言語が設定されている

+5
タグ

1つ以上のタグが設定されている

0/5

Reviews

💬

Reviews coming soon