npx claudepluginhub phi-labs-ltd/claude --plugin developerUse this agent to produce a Quint formal specification from any input — user stories, a Linear/RFC ticket, a PDF, pseudocode, a Python prototype, or an existing implementation (Rust, Move, Solidity, etc.). The agent works from a protocol-designer mindset: it models what the protocol should do, not what the code currently does, and will deliberately diverge from the implementation when the implementation is missing an invariant, a guard, or an error branch. Trigger when the user asks to "write a spec", "formalize this", "draft a Quint spec", "specify this protocol", or asks for help designing invariants. Also trigger proactively when the user is authoring or modifying a smart contract on Sui (Move), Solana (Rust programs), or Ethereum / EVM chains (Solidity, Vyper) — formal specification is the expected workflow for these contracts.
Use this agent to get an independent, critical second opinion on pending code changes — reviewed fresh, without the baggage, assumptions, or sunk-cost bias of the original implementation session. The agent reads diffs with the eye of a skeptical senior developer: scope creep, unnecessary abstractions, and large changesets must justify themselves. Trigger when the user asks for a "second opinion", "sanity check", "sense check", "review my changes", "is this the right approach", or before shipping a non-trivial diff. Also trigger proactively after completing a large or architecturally significant change.
This skill should be used when the user asks to "formalize" a requirement, "convert this into spec", "write a Quint spec", "specify X formally", "translate this into Quint", or wants to produce a Quint model from informal inputs (user stories, PDFs, pseudocode, Python prototype, or existing implementation code such as Rust). Also AUTO-TRIGGER proactively whenever the user is writing, modifying, or reviewing a smart contract on Sui (Move — `.move` files, `Move.toml`), Solana (Rust programs — `solana_program`, `anchor_lang`, `#[program]` modules), or Ethereum / EVM chains (Solidity — `.sol` files, Vyper — `.vy` files) — formal specification in Quint is the expected workflow for these contracts and the skill should surface even when formalization is not explicitly requested. Guides a three-phase workflow — draft spec, negotiate invariants with the user, write comprehensive tests — targeting the Quint language (https://quint.sh).
Engineering discipline for writing, editing, and refactoring code — think before coding, keep it simple, make surgical changes, and execute against verifiable goals. Apply whenever the task involves producing or modifying code (features, fixes, refactors), especially non-trivial changes where over-engineering, scope creep, or ambiguous success criteria are likely. Skip for purely exploratory questions, documentation-only edits, or throwaway one-liners. Derived from [Andrej Karpathy's observations](https://x.com/karpathy/status/2015883857489522876) on LLM coding pitfalls
Claude Code plugins by and for Phi Labs Ltd.
| Plugin | Ships | Description |
|---|---|---|
personalities | Output styles | Different voices for Claude. Currently: Crypto Bro. |
developer | Agents, skills | Developer tooling. Currently: the second-set-of-eyes review agent, the protocol-formalizer Quint-spec agent, the karpathy engineering-discipline skill, and the FormalSpecification Quint-spec skill. |
bolt-theme | Theme | A dark color theme with Bolt Liquidity's signature yellow accent. |
Add the marketplace once:
/plugin marketplace add phi-labs-ltd/claude
Then install any plugin you want:
/plugin install personalities@phi-labs-ltd
/plugin install developer@phi-labs-ltd
/plugin install bolt-theme@phi-labs-ltd
Run /reload-plugins after installing for it to take effect.
personalitiesPick a voice with /config → Output style → personalities:crypto-bro. Revert with /config → Output style → Default.
developersecond-set-of-eyes — an independent-review agent that reads your pending diff with the critical eye of a senior engineer. Ask for a "second opinion", "sanity check", or "review my changes" and Claude will delegate to it. Read-only: it cannot edit, stage, or commit.protocol-formalizer — an agent that drafts Quint specifications from user stories, tickets, PDFs, pseudocode, or existing implementations. Works from a protocol-designer mindset: models what the protocol should do, not what the code currently does, and flags spec-vs-code divergences (e.g., invariants the implementation is missing). Delegated to on "write a spec", "formalize this", "draft a Quint spec", or proactively when editing smart contracts on Sui / Solana / Ethereum.karpathy — an engineering-discipline skill (think before coding, simplicity first, surgical changes, goal-driven execution). Loads automatically on non-trivial code tasks.formal-specification — a skill that guides a three-phase workflow for producing a Quint spec: draft the model, negotiate invariants collaboratively, then write comprehensive deterministic tests. Includes a Quint syntax cheatsheet, house-style guide, invariant-pattern catalogue, and quint-connect trace-replay integration notes. The protocol-formalizer agent uses this skill; it also loads directly when you ask to "formalize" or "convert this into a spec".bolt-themeRun /theme and pick Bolt Liquidity — a dark base with Bolt's signature yellow (#FFED2C) for Claude's voice.

Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.
Uses power tools
Uses Bash, Write, or Edit tools
Orchestrate multi-agent teams for parallel code review, hypothesis-driven debugging, and coordinated feature development using Claude Code's Agent Teams
Complete developer toolkit for Claude Code
Comprehensive toolkit for developing Claude Code plugins. Includes 7 expert skills covering hooks, MCP integration, commands, agents, and best practices. AI-assisted plugin creation and validation.
Comprehensive .NET development skills for modern C#, ASP.NET, MAUI, Blazor, Aspire, EF Core, Native AOT, testing, security, performance optimization, CI/CD, and cloud-native applications
AI-powered development tools for code review, research, design, and workflow automation.