Plugins

Canon-grounded coding-discipline workflow (TDD/brainstorm/SDD/review) for Claude Code + Codex.

Key-free cross-modal (GUI/TUI/CLI) UX design generator — emits portable DESIGN.md + ui-flows.md.

Deterministic workflow driver conducting the loom principles→design→spec→code pipeline across station plugins.

Key-free generator for a PRINCIPLES.md project constitution — north star + falsifiable Product Principles, plus optional Design/Engineering Principles.

Key-free GENERATE-layer spec skills — expand a seed into a high-recall spec + critique omissions.

Loop engineering wizard for Claude Code. Answer 2 questions and Claude orchestrates a fully autonomous 7-agent team (tool-scout, researcher, developer, qa-tester, verifier, auditor, memory-keeper) until your goal is met.

Diagnose a repetitive task into one of 5 loop shapes (Batch/Pipeline/Refine/Watch/Explore), interview for the blanks, and auto-inject an independent verifier (maker ≠ checker) and a hardstop, then stamp it as a reusable slash command.

Stack-agnostic loop-engineering harness: pwsh hooks (session context, Bash/PowerShell deny-list guard, Stop+SubagentStop gate on build/verify.ps1 with a real round-trip cap and ledger logging), maker/checker/explorer/verifier sub-agents with real effort routing, /init-harness bootstrap, /work-next automation, per-stack verify templates that fail loudly until wired, and harness-audit/claude-code-hook-facts/verify-gate-authoring skills.

Guardrails, verification patterns, goal-condition templates, and loop discipline for Claude Code loops. Complements ralph-loop with the discipline that makes loops trustworthy.

Turn any task description into a ready-to-paste six-part agentic loop prompt. /loopgen <your task>.

Meta-orchestration skill: picks an execution strategy (single pass, bounded loop, or parallel sub-agent swarm) and drives it to completion with durable state and audit trails.

from srnnkls/tropos

Capture tribal knowledge as a byproduct of normal work and serve it back cited, work-scoped, and staleness-checked.

Explicit context management for multi-agent systems — decisions, patterns, failures

LLM-optimized knowledge graph: session notes, auto-extracted knowledge, pluggable briefings, magic context injection.

Mine your SpecStory coding histories into a persistent corpus and forge your repeated workflows into installable agent skills.

Deterministic epic/PRD/TDD workflow skills for AI-coding teams (13 skills: orient, grilling, to-epic, orchestrate-epic, tdd, debug, document-epic, …). Depends on lore.

Reads scoped lessons at session start and nudges a retrospective when a tool fails, backed by the LoreKit MCP server. Requires LOREKIT_MCP_URL and LOREKIT_TOKEN in the environment (or a project .mcp.json written by `npx @lorekit/cli install`).

Skills for daily notes, project closing, and YouTube summaries

Generate professional Lottie animations from static SVGs. Replaces After Effects for motion graphics.

Read DOM marks placed in the browser as precise frontend tasks.

Build, iterate on, deploy, and manage Lovable apps from Claude Code. Bundles the official Lovable MCP server (remote, OAuth 2.1) and adds focused commands for the common build/iterate/database workflows, with credit- and publish-safety prompts.

from bennoloeffler/maude-claude-vunds-plugins

from LovelyStack/lovely-dev

Helps legalquants/lq-ai maintainers triage PRs, issues, and dependabot traffic against lq-ai's canon. Recommends lanes, drafts receipts and responses. A human decides, every time.

Persistent, team-shared memory for AI coding agents

Little Shit Done — lightweight Spec-Driven Development for Claude Code. Plan small, execute in isolation, ship fast.

Price bonds, analyze yield curves, evaluate FX carry trades, value options, and build macro dashboards using LSEG financial data and analytics.

Language Server Protocol specialist building unified code intelligence systems through LSP client orchestration and semantic indexing

LSP 实现、代码智能与语义索引专家