Productivity plugins

Auto-convert structured JSON to TOON only when token savings are meaningful.

2 skills

Codex workflow guidance for worktree-fleet local coordination.

2 skills

Plain Language mode: short sentences, active voice, no AI-isms. Includes the /plainlang:scrub skill for rewriting existing prose.

1 skill · hooks

Cuts noise from agent communication. Maintains an auditable model of what you already know, and uses it to decide what to leave out.

1 skill · hooks

Recommends the right Claude model (Haiku/Sonnet/Opus) and effort level at session start by analyzing your first message, then pauses so you can switch before work begins.

1 command · hooks

Ryvn tools for Claude Code - manage environments, services, deployments, and infrastructure

1 skill · hooks

Coordinate tracked work with taskix and read-only Markdown boards.

1 skill · hooks

Chaotic, manic, slightly insulting responses. Technically accurate. Emotionally questionable.

1 skill · hooks

Blocks the user from resuming a conversation after the Anthropic prompt cache TTL (5 minutes) has expired, preventing expensive cache-resurrection writes.

1 command · hooks

Weighs everything your coding agents load at startup, prices it in tokens from your own transcripts, and parks the dead weight with a receipt and an undo.

2 skills

100-word cap, one framed ONE THING per reply. Runs last, wins every conflict.

1 skill · 1 command

Inject a character personality into every conversation via a SessionStart hook. Switch personalities at any time with /claude-code-personalities:personality.

1 skill · hooks

Agentic hardware co-design from brief to KiCad, firmware, 3D review, and explicit fabrication blockers.

1 skill · 1 MCP server

Fetch Teams meeting transcripts via a Microsoft 365 MCP connector and save them as clean local markdown. Includes an in-progress-meeting guard.

2 skills

LinkedIn API toolkit — MCP server and workflow guidance for Claude Code

1 skill · 1 MCP server

Forces a Hypothesis Ledger with a mandatory non-local (zoom-out) hypothesis, 5 Whys with evidence gates, a systems lens, and information-gain gates before any bug fix, root-cause diagnosis, or plan is declared complete.

1 skill · hooks

Persistent memory with knowledge graph, importance scoring, compaction recovery, and Memory Protocol. Gives Claude a brain that survives across sessions.

1 skill · hooks

把笔记 / 会议纪要 / 工单按 12 题标号收进带出处的账本，AI 做判断前先读它。挂不上题的一个字都进不来。含周报（指出哪一段还是空的）和零内容的可截图进度卡。数据只在本机，不联网不上传。

1 skill · hooks

Lightweight token usage display for Claude Code status line. Zero LLM calls — pure bash + jq.

1 skill · 1 command

Provides the /llm-wiki-init:wiki-init and /llm-wiki-init:wiki-status commands for scaffolding and inspecting a persistent, LLM-maintained knowledge base on any topic.

2 commands

Agentic technical project/product manager — researches a problem's evidence, frames an approval-gated epic with hypothesis and success metrics, decomposes it into a plan (milestones, risks, MoSCoW) and INVEST tasks with binary acceptance criteria under docs/pm/, and tracks status across sessions with a persistent epic index and product memo

1 skill · 1 command

Reversible file changes with automatic recovery cleanup, separate document/media budgets, protected media replacement and keep-both/cancel, sync-aware safety, and policy-driven command/content blocking.

1 skill · hooks

Explicit-only general architecture standards lookup

1 skill · hooks

Systematic maintenance workflow for open MicroPython PRs: recon, rebase, CI, push

1 skill · 1 command

List, sort, and triage open MicroPython PRs with branch and worktree status

1 skill · 1 command

Auto-name your Claude Code sessions. A Stop hook watches your conversation, scores work-density deterministically, and asks Haiku for a `domain: clauses` title when meaningful work has accumulated. Seven /smart-rename subcommands give you manual override, anchoring, freeze/force, and inspection.

1 skill · hooks

Answers retrospective questions from the recorded session and closes work with the shared handoff receipt.

1 skill · hooks

Keeps coding work scoped to the requested outcome and enforces the shared three-loop stop rule with a stateful PostToolUse hook.

1 skill · hooks

Evidence-backed engineering discipline: acceptance contracts, root-cause fixes, a hook-enforced verification gate, and the shared three-loop stop rule.

1 skill · hooks

Export Claude Code and Codex conversations to terminal-styled PNG, SVG, PDF or HTML.

1 skill · 1 command