S40911120
精选
recensa
Recensa is a self-hosted web viewer that indexes Claude Code session transcripts into a local SQLite database, enabling full-text search, replay, and audit of all past agent conversations without uploading data anywhere.
AI 仓库情报摘要
FR-AI / ANALYSIS为什么值得关注
它引入了基于合同优先的结构化方法，内置测试、验证和自我修复，旨在生成比典型 AI 代理更可靠、更完整的代码。
适合谁使用
- AI辅助软件开发者
- 探索自主开发的工程团队
- 处理遗留代码的开发者
- 关注代码质量需要严格验证的团队
典型使用场景
- 快速构建原型并确保测试覆盖
- 逆向工程并扩展现有代码库
- 从需求到 PR 的全自动化功能开发
- 对 AI 生成代码实施质量门控
项目优势
- 每个里程碑都强制实施测试驱动开发
- 独立任务和验证器的并行执行提升速度
- 由不同代理对代码进行对抗性验证
- 在规划前对现有仓库进行逆向工程（棕地支持）
使用前须知
- 需要 Kiro 代理基础设施（superpowers + gstack）
- 实验性项目，未经外部安全审查
- 生成的代码在生产使用前需经过安全审查
- 浏览器验证依赖 Playwright MCP
README 快速开始
SpecShip
A Kiro Power autonomous engineering workflow that orchestrates AI agents through a complete recon → plan → build → validate → ship pipeline — with brownfield reverse engineering, TDD, parallel execution, adversarial validation, and spec-driven quality gates.
Architecture
Editable source: docs/architecture-diagram.svg
The 4 Phases (+ Brownfield Recon)
| Phase | What happens | Output |
|---|---|---|
| RECON (brownfield only) | Reverse engineer existing code before planning: repo map, technology stack, components, APIs, data models, business flows, baseline tests, preserved behavior, and change impact. | .specship/.../artifacts/reverse-engineering/ |
| PLAN | Brainstorm → market research (real web search, minimum 5 queries) → sprint contract → pre-generate test cases → implementation plan. Each step feeds the next. | requirements.md + design.md + tasks.md + artifacts |
| BUILD | Execute tasks.md milestone-by-milestone. Parallel batches for independent tasks. Every behavior follows Test-Driven Development (write failing test → watch it fail → implement → pass → commit). Hard gate blocks each milestone until typecheck + tests + build pass. | Tested, committed code per milestone |
| VALIDATE | Up to 7 adversarial validators run in parallel as independent subagents, each delegating to its gstack specialist skill (code→review, security→cso, browser→qa-only). Code, Security, Integration, Browser, Design, Alignment (always) + Load (if performance NFR exists). Each produces a typed verdict with evidence. Aggregate decides: merge / recover / escalate. | Typed verdicts → merge or recover |
| SHIP | PR + changelog + archive. Ship report with timing, bugs caught, milestone breakdown. If no git remote, prepares locally with push instructions. | PR ready to merge |
Why SpecShip?
AI agents build fast but shallow — working skeleton in 7 minutes, 60% of features missing, no tests, no error handling. SpecShip fixes this:
| Principle | What it means |
|---|---|
| Test-Driven Development | No production code without a failing test first. The build gate blocks every milestone until typecheck + tests + build all pass. Catches bugs within minutes, not hours. |
| Parallel execution | Independent tasks run simultaneously. A |
项目描述
Spec-driven autonomous engineering workflow for AI coding agents: recon → plan → build → validate → ship — with TDD, adversarial validation, and anti-slop quality gates. Packaged as a Kiro Power.
相关仓库与替代方案
根据分类、Topic 和编程语言匹配的相似项目。
bjan
精选
claudemux
Claudemux is a lightweight, dependency-minimal message bus that lets multiple Claude Code sessions communicate in real-time over tmux, enabling one session to ask another a question and receive the answer without manual intervention.
mereyabdenbekuly-ctrl
精选
clodex-ide
Clodex is an open-source, local-first agentic IDE that combines persistent AI tasks, code, terminal, browser, Git, models, memory, and governed execution in one Electron workspace, currently in technical preview.