subagents plugins

by Marat Karimov

Catalyst: disciplined agentic development - SDD pipeline with decision-boundary model tiering, spec hardening, premortem gates, verification, and handoffs.

by Francesco Sorrentino

Keeps Claude Code from spending your quota on work the top model didn't need to do: declared pre-budgets with deterministic 2x/3x enforcement, cheap-executor routing, objective token telemetry. Small one-offs pay a 5-20% premium (measured) - built for sessions that delegate. New in 1.42: reads video and audio for you at zero Claude tokens - a whole clip in one contact-sheet image, transcription on your CPU, voice-over timing checked before the studio; long footage on free-tier Gemini.

by Harry Yep

Multi-model orchestration: tier agents, routing rules, blind panel, ultra fleet, model config, and drift doctor.

by ken2403

Claude Code Higher Agents for building ONE feature thoroughly: deep red-teamed plan to human approval, per-task implementation review, a risk-scaled pre-PR gate, independent PR review, rigorous feedback handling, gated merge, conflict resolution, and safe worktree cleanup. Requires the superpowers plugin; model-agnostic.

by Lauren Tan

if you want to go fast, go deep first. pstack helps you write less, but higher quality code. rigorous agent workflows you can parallelize with confidence. Ported from cursor/plugins.

by ken2403

Command-free, skills + subagents plugin for fast single-feature development: digest a plan, get approval, isolate in a worktree, implement, and open a PR — Sonnet build / Opus review, with graded effort.

by Leo Liang

Adaptive delegation, cost-aware model routing, and deferred workflows across coding harnesses.

by Paulo Soares

Spec-driven development framework — hybrid orchestration with 8 real subagents (executor, advisor, simplifier, code-reviewer, QA, adversarial verifier, researcher, roadmap), per-task model routing by complexity, curated project memory graph feeding every spawn, opt-in parallel waves with worktree isolation, /pwdev-code:fleet for running multiple phases end-to-end (plan→execute→review→verify) unattended in isolated worktree+docker+tmux sessions, external CLI delegation (codex, opencode, kimi, gemini, kiro + smart /delegate), optional external reviewer CLI, strict dual-verifier mode, 6 phases with correction loops, deterministic SQLite audit via hooks, and extensible skills that turn requirements into verified, traceable code.

by Paulo Soares

Framework de copy genérico e treinável — um arquivo de contexto define marca, ICP e voz, e 20 skills cobrem o ciclo completo: pesquisa VOC, brand voice, criação (landing, social, ganchos, reaproveitamento), revisão em 7 sweeps com anti-slop, e camada de análise que fecha o loop (perf-analyzer, perf-patterns, perf-optimize). Inclui 5 subagentes reais e revisão adversarial de conversão.

by Paulo Soares

Simplified AI-assisted feature development — PWDEVIA 7-question action plans created inline (interviewing you, max 2 rounds), executed by a real executor subagent with fresh context and IMPLEMENT/REPORT modes, an advisor subagent (strong model) for hard mid-task decisions, read-only consumption of pwdev-code's curated project memory when present, deterministic SQLite audit via hooks. Fast, practical, no ceremony.

by Paulo Soares

Stack-agnostic UI/UX engineering framework for Claude Code — 6 real subagents + inline orchestrator/theme personas, 5-phase workflow with human gates, Figma integration, WCAG 2.1 AA, deterministic SQLite audit via hooks. Supports shadcn-vue, shadcn/ui, PrimeVue, Untitled UI, Tailwind Plus, and custom stacks.

by Maxwell Volz

slashwork earner. /earn init authenticates and scaffolds an earner agent folder with run settings (model, bypass_permissions, default_duration); a bare /earn runs the default duration, /earn <goal> holds the live task feed over SSE, claims offloaded subagent tasks the moment they land, runs each with the folder's configured agent, and a SubagentStop hook submits the artifact with token usage. Earn credits and a per-class score for accepted work.

by Maxwell Volz

slashwork offloader. Interception is on by default: /work init authenticates (a bare /work with no token does the same), and from then on a PreToolUse hook routes self-contained subagent spawns to the offload network and returns the artifact in place, falling back to the local spawn on any miss. /work off pauses routing per project, /work on resumes it; a bare /work shows status and the dashboard.

by Mark Summer

Drive GitHub issues through planner → implementer → verifier subagents, locally on a Claude Code subscription. Steady-state cycle skill (issue-cycle), persistent verification baseline, policy-gated plan auto-approval, bounded CI auto-fix, crash recovery with WIP checkpointing and resume (not restart), backoff/retry on transient dispatch failures, a per-issue dispatch ledger with pre-advance checks and mechanical closing reconciliation, per-stage status reporting, greenfield kickoff (project-kickoff), worktree-parallel implementation with a supervisor loop, label lifecycle, repo onboarding (harness-setup), merges human-gated by default with opt-in policy-gated merge autonomy guarded by base-branch assertion, loud permission-denied escalation, post-merge verification, and an opt-in test-suite ratchet that files evidence-backed, test-only coverage issues under a CLAUDE.md policy.

by Michael Kissinger

Spec-anchored workflow for autonomous coding runs on GPT-6 Astra in Codex: brief, run, verify, checkpoint. astra mode has Astra implement each slice; astra-crew routes implementation to gpt-5.6-sol or gpt-5.6-luna subagents by task class. Same spec contract as the Claude Code plugin in this repo.

by johantor

Team-style feature delivery for Claude Code: a captain agent plans the work and delegates each step to backend, frontend, test, and design specialists behind hook-enforced guardrails, with a consolidated review gate before anything ships.

by Dan Jones

Dan's personal Claude Code skills and automations. Multi-agent planning with a review panel on the model you choose, and the prompts I reach for every day, consolidated.

by Patina Project

Patina Project's owned engineering workflows, forked from pstack.

by cjthompson

Reduced-toolset sub-agent profiles (read-only, lean-executor, standard-executor, main, full-executor) that lower System-tools token overhead vs. spawning the default agent. Pair with the project-tasks skill, which dispatches scouts, executors, and verifiers to these agents.

by Jordan Richlen

Two reusable multi-agent orchestration templates for research-and-verify work on Claude Code's Workflow tool: fan out research over dimensions, then adversarially verify the claims that research surfaces so plausible-but-wrong findings don't survive. Ships a shared skill, two workflow-script templates (derived-verify, pipelined-verdict-wins), and a worked example.

by Vivswan Shah

Run a session as an orchestrator: parallel worktree subagents, gated landings (direct or PR-based), fleet monitoring.

by David Pham

Pasture agents for Claude Code — the five protocol roles (supervisor, worker, reviewer, architect, epoch) as installable subagents, generated from the pasture Go codegen.

by cjthompson

Capture project tasks inline, store them in SQLite, group work under plans, dispatch work, and generate changelogs.

by Abhirup Banerjee

Delivery discipline for AI coding agents. Writes the five-layer Stackwich policy (Prompt / Context / Harness / Loop / Graph) into CLAUDE.md and scaffolds advisor, executor and verifier subagents.