Development plugins

Lets a coding agent ask what its Claude Code harness costs: hook time, text hooks inject into context, and Hindsight memory's share, with the plugin or settings file behind each cost. Read-only MCP server; needs Bun

1 MCP server

Keep each T3 Code thread's git commits attributed to the human operator it works for.

hooks

Agent skills for measuring how Claude Code agents perform, starting with what they cost

1 skill

Have a different AI model review the code an agent wrote, before it reports the work as done. Claude Code asks Codex CLI.

1 skill

Encontra tempo de runner desperdiçado em GitHub Actions e Gitea Actions e prepara pequenos patches YAML revisáveis. Fork do dsh-ci-diet com suporte a Gitea.

1 skill

Enforces the architecture a repository declares in .claude/architecture.json at Edit/Write (layer direction, forbidden imports), notes code smells with the refactoring to apply, and ships a refactoring.guru-keyed catalog and /arch audit.

1 skill

Architects modular SwiftUI apps as one Swift package per feature, split into domain, UI and live-service targets over a small shared package, with an app target that composes screens, injects services and owns navigation. Use when adding a feature or screen to a SwiftUI app, modularizing an iOS app into Swift packages or SPM modules, drawing module boundaries or dependency direction, designing domain models, services or MVVM view models (loading, errors, pagination, search), wiring navigation, routers, coordinators or dependency injection between features, deciding where a type belongs or what to name it, or reviewing an app's architecture, even when the user doesn't name the architecture.

1 skill

Self-check a Codyssey AI Native final project against the official rubric

1 skill

TypeScript 7 native LSP (tsgo) for Claude Code with pull-to-push diagnostics and idle stop

1 skill

Migrate a repo's Cursor setup to Claude Code while other agent tools keep working.

1 skill

Analysis-only PreToolUse hook that asks before edits that weaken tests.

hooks

Types-first, closures-as-ports Rust style guide for agents.

1 skill

A single master engineering skill with progressive, on-demand technical references.

1 skill

Every data-driven UI ships with its loading, empty, error, overflow, and offline states.

1 skill

Rate any webpage against the Search Quality Rater Guidelines with cited E-E-A-T, Page Quality and Needs Met ratings.

1 skill

Programming the Epson TM-U220 9-pin dot matrix receipt printer over ESC/POS: text styles, red, images, charts and stored logos.

1 skill

Senior-engineer PR review that writes paste-ready, line-anchored comments in plain English, verifies blockers with a second agent, and can post them to GitHub as one review.

1 skill

A skill that tells the agent which official feature, auto-instrumentation or measure-everything package to use for each measurement

1 skill

Diagnose and fix failed GitHub Actions runs (and GitLab, Jenkins or local logs) with redacted log digests, 160 known failure patterns and a consistent report.

1 skill

A symptom-first registry of the traps your project has hit, with stable IDs and a validator that keeps it honest.

1 skill

Run builds, tests and code generators in your own GitHub Codespaces, with jobs that survive the session and checked patches.

1 skill

Make your coding agent verify its own change before it calls the work done: ProofMerge writes probes aimed at what the change could break, runs them on base and head in Docker, and returns PASS, REVIEW or BLOCK with replayable evidence.

1 skill

Capture the pre-agent state and require causal evidence for the real patch before work is declared complete.

1 skill

Build, deploy and maintain Velk CMS sites: the site.json schema, themes and templates, installs and engine upgrades, and deploys, content edits and backups over SSH (rsync, mysqldump) to the servers you configure.

1 skill

Judge competing UI and design candidates with blind, order-swapped pairwise comparisons and a Bradley–Terry fit instead of saturating 0–10 scores.

1 skill

Senior software-engineering partner for implementing, reviewing, assessing, refactoring and securing code — including architecture, design patterns, database migrations, test strategy and threat modelling — with a traceable rule catalog and evidence-labelled findings.

1 skill

One shared, hidden GPU browser for every Claude Code agent: screenshots, animation frames, frame-time statistics and downloads through a job queue.

1 skill

The annokit-setup skill: add AnnoKit to an Expo or bare React Native app and start the stub server that hands tasks to your coding agent.

1 skill

Reconstruct the publicly served frontend of a website locally and prove fidelity by running identical probes against original and clone.

1 skill

Teaches Claude Code the browser CLI (npm install -g @phnx-labs/browser-cli): open signed-in sites, navigate, click, type, read pages and take screenshots.

1 skill