Help us improve
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
By alexanderop
AFK coding pipeline: spec → vertical slices → TDD implementation loops → refactor pass → agentic QA → multi-agent review. Human judgment at the edges, agent execution in the middle.
npx claudepluginhub alexanderop/afk --plugin afkReviews a branch diff for correctness bugs, weak tests, and structural problems. The workhorse reviewer, dispatched by afk:review at every tier above trivial.
Reviews a branch diff for documentation drift — stale AGENTS.md/CLAUDE.md, READMEs that now lie, missing notes for new env vars or commands. Dispatched by afk:review at lite tier and above.
Implements ONE vertical slice ticket with forced TDD (red-green-refactor). Dispatched by afk:ralph with the full ticket text, check commands, and branch — never self-invoked.
Reviews a branch diff for measurable performance regressions — N+1 queries, unbounded loops over user data, accidental re-render storms. Dispatched by afk:review for full-tier reviews.
Reviews a branch diff for exploitable or concretely dangerous security issues. Dispatched by afk:review for full-tier reviews and whenever security-sensitive files change.
Use when the user hands over a big feature (a spec, a PRD, or a 5+ point request) and wants it shipped end to end — orchestrates the full AFK pipeline: spec → slice → implement → refactor → QA → review → PR, with human gates only at the edges.
Use when implementation and refactoring are done but no human has seen the feature run — exercises the real running project on whatever surface it actually has (browser via agent-browser, HTTP API via curl, or CLI invocations), walks the happy path and negative paths from the spec, and writes an evidence-backed QA report.
Use when slice tickets exist and it's time to implement — runs a fresh-context subagent loop per slice with forced TDD (red-green-refactor), then verifies each slice against its ticket with an independent reviewer before moving on.
Use after all slices of a feature are implemented and before QA/review — a dedicated cleanup pass over the branch that collapses duplication across slices, removes dead code, and fixes naming, because the implementation loops will not do this on their own.
Use after a pipeline run, a correction from the user, or any session with notable friction — captures learnings into structural enforcement (lint rules, AGENTS.md) first, and into the project's .afk/brain/ memory only when a rule can't encode it.
Modifies files
Hook triggers on file write and edit operations
Uses power tools
Uses Bash, Write, or Edit tools
Share bugs, ideas, or general feedback.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
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.
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.
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.
Complete developer toolkit for Claude Code
Intelligent draw.io diagramming plugin with AI-powered diagram generation, multi-platform embedding (GitHub, Confluence, Azure DevOps, Notion, Teams, Harness), conditional formatting, live data binding, and MCP server integration for programmatic diagram creation and management.
Create skills, subagents, hooks, commands, and plugins for Claude Code
Modern Vue 3 development with TypeScript, Composition API, defineModel bindings, Testing Library for user-behavior tests, and MSW for API mocking
Vue language server for Claude Code
No description provided.
A Claude Code plugin that makes the AFK coding pipeline the path of least resistance: human judgment at the edges, agent execution in the middle. You align on the spec, the agents ship it, you review the PR.
Built for developers who haven't yet found a good workflow with AI coding agents. Install it, and the failure modes that ruin big AI-built features — context overflow, skipped tests, never-happening refactors, unreviewed slop — get caught by the pipeline instead of by your users.
1. Align on spec ── HITL ── AI interviews YOU, writes the PRD
2. Slice the ticket ── HITL ── vertical slices, you approve the cut
3. Implement per slice ── AFK ── fresh-context TDD loop + spec review per slice
4. Refactor pass ── AFK ── the step LLMs always skip
5. Agentic QA ── AFK ── drives the real surface: browser, API, or CLI
6. Review ── AFK ── risk-tiered multi-agent review
7. PR + handoff ── HITL ── you review, business does UAT
Small tickets (1–3 points) skip all of this — the plugin's session-start bootstrap teaches Claude to size the ticket first and only reach for the pipeline at 5 points and up.
Installation differs by harness: Claude Code or GitHub Copilot CLI. If you use both, install afk separately for each one.
Prerequisite: Claude Code installed and authenticated. If /plugin doesn't exist in your session, update Claude Code first.
1. Add this repo as a plugin marketplace (one-time, inside any Claude Code session):
/plugin marketplace add alexanderop/afk
2. Install the plugin:
/plugin install afk@afk
3. Verify it's loaded — run /help; you should see the skills listed under the afk: namespace. The session-start bootstrap (ticket sizing + skill routing) is active from your next session onward.
4. First run in a project:
/afk:setup
Updates ship via the marketplace: /plugin marketplace update afk.
Copilot CLI reads the same plugin format. In your terminal:
1. Add this repo as a plugin marketplace (one-time):
copilot plugin marketplace add alexanderop/afk
2. Install the plugin:
copilot plugin install afk@afk
3. Verify it's loaded:
copilot plugin list
Skills written for afk use Claude Code tool names; Copilot CLI maps them automatically via skills/using-afk/references/copilot-tools.md. Update later with copilot plugin update afk.
The plugin's hooks emit Copilot's output format too, but Copilot CLI has a known bug where plugin-shipped hooks are listed at startup yet never execute (copilot-cli#2540). Two backstops cover this: /afk:setup embeds the ticket-sizing gate in your project's CLAUDE.md (which Copilot reads in every session), and you can load the using-afk skill manually — type / and pick it from the skill list — to inject the full bootstrap.
git clone https://github.com/alexanderop/afk
claude --plugin-dir ./afk
This loads the plugin for that session only — useful for kicking the tires or hacking on the plugin itself (edit + /reload-plugins to pick up changes).
In any project:
/afk:setup # one-time: backpressure audit (tests, lint, types)
/afk:pipeline build a multi-step booking wizard
Claude interviews you for the spec, proposes vertical slices, and after your approval runs implementation, refactoring, QA, and review without you — then opens a PR with the QA and review reports attached.
| Skill | What it does |
|---|---|
/afk:setup | Audits & fixes backpressure, records the project's QA surface, crafts a lean WHAT/WHY/HOW CLAUDE.md, stands up the .afk/brain/ knowledge vault |
/afk:spec | Interview-mode PRD — AI asks, you answer, one question at a time |
/afk:slice | PRD → vertical slice tickets (UI + API + test, each shippable) |
/afk:ralph | Fresh-context TDD subagent loop per slice, with independent spec review |
/afk:refactor-pass | Dedicated cleanup pass: duplication, dead code, type holes |
/afk:qa | Walks happy + negative paths from the spec on the project's real surface — agent-browser for UIs, curl for pure APIs, real invocations for CLIs — evidence-backed report |
/afk:review | Risk-tiered review: security/quality/performance/docs specialists + judge pass |
/afk:pipeline | The meta-skill — runs all of the above end to end |
/afk:reflect | Banks learnings: lint rule → script → CLAUDE.md → .afk/brain/ note → skip |
Both extension points live in your repo, versioned with the code and shared by everyone on the team — no plugin fork, no reinstall.