Development plugins

Installs the screencast-mcp server plus skills for bug-repro GIFs, README demos, and verifying UI changes on screen.

3 skills

Skills for working with Kerb: refusals and hints, a debugging loop over Kerb's records, and a verification check before finishing.

3 skills

Token optimization for Claude Code: audits session transcripts, applies low-risk reversible fixes, recommends model, workflow and instruction changes with evidence, guards against retry loops and writes handoff files.

2 skills · hooks

Lean 4 and Mathlib workflows: developing and refactoring proofs, searching Mathlib, driving Lake builds, and auditing whether statements and the prose citing them mean what they claim.

3 skills

Two AI agents (Architect and Reviewer) collaborate through iterative refinement until consensus on optimal design

1 skill · 1 command

Design systems on a shared visual canvas.

1 skill · 1 MCP server

Capture evidence of corrections and failures, then curate reusable learning episodes and eval candidates.

1 skill · hooks

Runs this repository's CI gates locally (lint, typecheck, tests, build, DB checks) and reports a compact pass/fail summary.

1 skill · 1 command

Collection of project continuity capabilities across sessions.

1 skill · hooks

Generate codebase slices (slice-codebase skill + codebase-slicer agent) and track doc-code staleness via the slice CLI.

1 skill · 1 agent

Scaffold large implementation tasks for iterative execution with Claude in a bash loop

2 skills

Skills for Informer development. Includes magic-apps (Informer App development with Vite, /informer:magic-apps) and marketplace-publishing (tag-driven release/publish workflow, /informer:marketplace-publishing).

2 skills

Build and query repository code indexes through Claude Code skills and MCP tools.

1 skill · 1 MCP server

Custom skills collection: A2A protocol development and EvoPalantir forecast runtime.

2 skills

Delegate tasks to Gemini CLI as a subagent, with Claude reviewing and presenting results

1 command · 1 agent

Rust engineering guidance for Halter.

2 skills

Create, revise, and inspect 3D CAD models of physical parts with Meshix.

1 skill · 1 MCP server

GPU-accelerated semantic code search and code intelligence. Find code by meaning, lookup symbols, query structure.

1 skill · hooks

Automated experiment loop for continuous improvement — bootstrap, benchmark, iterate, log. Port of karpathy/autoresearch for Claude Code.

1 skill · 1 command

Give your text-only agent eyes: analyze_image MCP tool, vision skill, and auto-loop clipboard hook. Bring your own multimodal model (mimo, Claude, Gemini, GPT-4o).

1 skill · hooks

Python language server based on basedpyright

hooks · LSP

Two-step repository assessment workflow: build repo context, then assess the repository against its real purpose.

2 commands

The Spacta skill (the ten Laws, where each kind of code goes, how to close a write's round trip, how to check behaviour) and a Stop hook that holds a turn which changed governed source and left `verify` red. Pairs with the `spacta` npm package, which carries the engine and the verifier the hook runs.

1 skill · hooks

Create and improve autonomous coding-agent prompts: context, done criteria, constraints with reasons, and harness checks

1 skill · 1 command

Pre-release multi-language impact, native enforcement, secure sandbox, reversible compression, durable sessions and paired benchmarking

2 skills

Starts from a blank folder or inspects an existing repository, interviews unresolved product and engineering decisions, and safely installs a project-specific Claude Code harness.

2 skills

Sharpitect MCP server with skills for intelligent C# codebase navigation. Automatically prefers semantic analysis over text-based searching.

1 skill · 1 MCP server

Agentic loop skill for building SaaS, DaaS, and AIaaS projects. Prompt -> Plan -> Execute -> Verify -> Iterate across the full stack, with Caveman token compression, frontend design, security review, and code review embedded. Bundles the agentic-project-builder skill.

2 skills

Two pull request review agents. the-nitcracker flags defects inline and writes a narrative walkthrough; pr-walkthrough writes orientation only. Both split output into defects, design forks, and implicit bets, and render judgment on defects alone.

2 agents

Build and maintain Reclaim Protocol data verification providers from your editor: capture a site's network traffic, synthesize a provider recipe, replay it, run a zkTLS proof, and publish it. Also inspects verification sessions and diagnoses failed ones. Use when verifying a data point about a user — a follower count, an account balance, an order history. Bundles the Reclaim MCP server, whose how_it_works tool serves the authoring guide one topic at a time.

1 skill · 1 MCP server