By ylxmf2005
Run a multi-agent software delivery pipeline: plan, implement, review, and verify code changes with specialized agents for code review, testing, security, performance, and acceptance, all coordinated through phased gates and evidence-based decisions.
扮演 AgentCorp 验收评审负责人:判断需求、实现、代码评审与验证证据是否足以支撑验收,并给出残余风险与发布条件。用于 AgentCorp 的验收评审(acceptance-review)阶段。
扮演 AgentCorp 对抗式评审者:挑战假设、挖掘失败模式、对需求与设计做压力测试,找出被忽视的风险,而不去重写方案。当 AgentCorp 中高风险、模糊、跨阶段或安全敏感的决策需要专职压力测试时使用。
扮演 AgentCorp 的 API Contract Reviewer:审查公共/共享 API 接口、schema、兼容性、对调用方的影响、auth 契约与错误语义。当 API 契约发生改动、需要专职 reviewer 把关时使用。
扮演 AgentCorp API 契约测试员:编写并真实运行测试,验证 API 是否符合其契约——request/response 形态、status code、auth/权限边界、错误语义、schema 一致性与兼容性。用于 AgentCorp 中聚焦 API 兼容性与契约行为的验证任务。
扮演 AgentCorp 逐 hunk 讲解走查者:当交付前需要逐块确认改动理由并部署本地 diff 走查页面时使用。
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
English | 简体中文
A multi-agent software-delivery pipeline packaged as Agent Skills. A Delivery Orchestrator classifies each task, routes every phase to a specialized role — planning, implementation, independent code review, and layered verification — and enforces explicit gates between phases.
Built on the Agent Skills SKILL.md standard, so the same
skills install in both Claude Code and Codex from one repository.
The orchestrator classifies each task into one of four paradigms —
dev/architecture-first, enhancement/delta-design, bugfix/hypothesis-driven, or
addition/simple — and runs it through a phased lifecycle in which the author of an
artifact never approves it.
Each paradigm runs a subset of the same phases. Review phases (red) and human gates
(orange) sit between work phases; request changes and reject return work to an
earlier phase.

Each phase has a dedicated skill. Review roles stay independent of the work they assess, and the orchestrator does not approve its own output.

Code-review findings are not fixed directly. An independent review-researcher
re-verifies each finding before any fix, filtering out false positives. Confirmed
fixes then run in parallel, partitioned by file ownership so no two workers modify the
same file.

Delegated phases move over assignment/receipt files. Each receipt is first checked mechanically (does the artifact exist; do the paths, author, and phase match) and only then judged against the phase's quality gate — the two checks are kept separate.

| Mode | Default | How it runs | When |
|---|---|---|---|
direct | no | No subagents at all; the orchestrator runs every phase itself, review phases produce drafts adjudicated by the sponsor at the human gate | small low-risk changes, or hosts without subagent support; requires explicit sponsor confirmation |
partial-delegation | yes | The orchestrator runs non-review phases itself; reviews are still delegated | regular, small-to-medium tasks |
full-delegation | no | Every phase is delegated to its owner via assignment/receipt | large or parallel work, or when independent authorship is required |
Every phase writes a Markdown artifact (with YAML frontmatter) to a fixed path, so a
completed task leaves a traceable, reviewable record. Delegated phases use an
assignment → receipt pair; each receipt is checked mechanically
(validate-handoff.py — does the artifact exist, do the paths/author/phase match)
before its quality gate, and manifest.md records every phase, owner, gate result,
and artifact path.
All artifacts live under teamspace/ in the work directory. This is local
coordination state and is never committed (add it to .git/info/exclude if it
appears); only the subdirectories a task needs are created.
npx claudepluginhub ylxmf2005/agentcorp --plugin agentcorpAstaNews — daily AI full-stack digest agent: papers, model releases, evals, infra/serving, MaaS changelogs, agents, embodied AI, safety, product/business, devtools. Community-extensible source registry.
Complete collection of battle-tested Claude Code configs from an Anthropic hackathon winner - agents, skills, hooks, and rules evolved over 10+ months of intensive daily use
Modern R development skills for Claude Code - tidyverse patterns, rlang metaprogramming, Bayesian inference, performance optimization, and more
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
20 SEO/GEO skills and 5 commands on one shared contract for keyword research, content creation, technical audits, schema markup, monitoring, quality gates, entity truth, and campaign memory.
Comprehensive SEO analysis plugin for Claude Code. 25 sub-skills (21 core + 1 orchestrator + 1 framework + 2 extension mirrors) and 18 sub-agents cover technical SEO, content quality, schema, sitemaps, Core Web Vitals, local SEO, backlinks, AI/GEO, ecommerce, hreflang, SXO, clustering, drift monitoring, and Google APIs. Includes optional MCP extensions, SPA-aware rendering, portability, and hardened SSRF/DNS-rebinding safe fetchers.
Develop, test, build, and deploy Godot 4.x games with Claude Code. Includes GdUnit4 testing, web/desktop exports, CI/CD pipelines, and deployment to Vercel/GitHub Pages/itch.io.