caro: fast Rust CLI that turns natural‑language tasks into a safe POSIX command. Built for macOS (MLX/Metal) with a built‑in model; supports vLLM/Ollama/LM Studio. JSON‑only output, safety checks, confirmation, multi‑step goals, devcontainer included.
Agent Claude Code
Use this agent to play a frustrated power-CLI-user beta tester for caro. Drives the daily 5 AM /caro.frustrated-qa routine. The agent runs short, ambiguous, real-world queries against the caro binary, captures every paper-cut (blanket replies, dropped intent, no streaming, no clarification), classifies findings by…
Agent Claude Code
Use this agent when working on release management, binary builds, distribution, or installation scripts for the Caro project. This agent has deep expertise in the multi-platform build system, GitHub Actions workflows, Cargo configuration, and all distribution channels. Context: User is debugging a failed release…
Agent Claude Code
Use this agent to deliver the Caro community waitlist feature (PR.
Agent Claude Code
Use this agent when the work touches Caro's brand identity — UI/UX audits, screenshot triage, communication with the claude.ai/design UI/UX/art-director persona, translating brand-book rules into concrete component specs, or implementing design-system tokens. Critically, ALWAYS spawn this agent for tasks that involve…
Agent Claude Code
Validates code changes against the project's consolidated knowledge and configuration rules (constitution). Use on git push to detect violations of agreed-upon standards like installation scripts, linking patterns, and configuration consistency.
Agent Claude Code
Daily creative query generator for caro. Seeds from CLI references, man pages, and Unix articles to produce novel natural-language queries; tests them against stable + branch builds; documents results; files GH issues for epic failures. Runs daily at 4am via .claude/automation/config/schedule.yaml. Examples — Context…
Agent Claude Code
Use this agent for multicultural holiday and celebration expertise, cultural sensitivity reviews, holiday theme validation, and adding new regional/religious holidays to the system. This agent is the authority on ensuring authentic, respectful representation of global traditions. Examples: Context: User wants to add a…
Agent Claude Code
Use this agent when you need to orchestrate complex software development projects involving multiple specialized domains, coordinate between different expert roles, or manage the full development lifecycle from specification to implementation. Examples: Context: User is starting a new complex software project that…
Agent Claude Code
Read-only adversarial reviewer that interrogates feature proposals, eval results, Hermes briefings, and PMF claims to counter AI confirmation bias. Pattern modeled on caro-frustrated-beta (assertive, evidence-driven, never blames the user) but pointed inward at our own claims rather than outward at the product.…
Agent Claude Code
Use this agent when you need to maintain project documentation, track development progress, manage releases, or update project status. Examples: Context: User has just merged a PR that implements the safety validation module and wants to update project documentation. user: 'I just finished implementing the safety…
Agent Claude Code
Use this agent when you need comprehensive developer experience strategy, CLI tool design, user onboarding flows, documentation architecture, or product management guidance for developer tools. Examples: Context: User is building a CLI tool and needs to design the complete user experience from installation to daily…
Agent Claude Code
Use this agent when you need code written in the pragmatic, systems-focused engineering style of Jeff Garzik. This agent should be invoked for:\n\n- Implementing low-level systems code (network protocols, file formats, database engines)\n- Building modular libraries with clear CLI wrappers\n- Writing…
Agent Claude Code
Use this agent to document leftover tasks as GitHub issues when ending a session with incomplete work. This agent maintains full correlation to the originating branch, PR, work plan, and user instructions. Context: User is ending a session with incomplete work user: "I'm done for today but the retry logic isn't…
Agent Claude Code
Use this agent when working on LLM integration aspects of the caro CLI tool, including model backend architecture, prompt engineering, inference optimization, and safety validation. Examples: Context: User is implementing a new model backend for caro. user: "I need to add support for Anthropic's Claude API to caro"…
Agent Claude Code
Use this agent when working on caro development that involves macOS/UNIX/POSIX systems integration, MLX framework implementation, shell command generation and validation, cross-platform compatibility, or native CLI tool optimization. Examples: Context: The user is implementing MLX bindings for Apple Silicon…
Agent Claude Code
Use when working on Caro fine-tuning, dataset curation, base-model evaluation, or training infrastructure. The agent is the persistent owner of the fine-tune pipeline targeting M4 Max 48GB unified memory. Examples — Context: user wants to add a new dataset collection hook. user: "We should log every embedded-backend…
Agent Claude Code
Use this agent when you need guidance on open source licensing, dependency compliance, attribution requirements, or legal implications of using third-party code in your project. This includes analyzing license compatibility, generating attribution files, assessing distribution requirements, and identifying potential…
Agent Claude Code
Use this agent when you need to design, implement, or refactor Rust CLI applications with a focus on open-source best practices, community-driven development, and production-grade system architecture. This agent excels at combining OSS advocacy with deep Rust CLI engineering expertise.\n\nExamples of when to use this…
Agent Claude Code
Read-only pragmatic-skeptic reviewer — the "lazy senior dev" who has been at the company longer than version control. Interrogates a change for over-engineering and over-strictness: code, abstractions, dependencies, files, or process steps that the problem does not actually need. The deliberate complement to…
Agent Claude Code
Use this agent when you need to implement complex software projects with a focus on pragmatic engineering decisions, MVP-first development, and production-ready architecture. This agent excels at breaking down ambitious projects into manageable phases while making smart technology choices that minimize technical debt.…
Agent Claude Code
Use this agent when you need to improve embedded LLM system prompt based on evaluation test failures. This agent should be used proactively when working on prompt engineering, LLM accuracy improvement, or when evaluation tests show poor command generation quality. Examples: Context: User wants to improve LLM command…
Agent Claude Code
Use this agent when you need comprehensive testing strategies, test implementation, or quality assurance for software projects. Examples: Context: The user has just implemented a new safety validation feature for their CLI tool and wants to ensure it's thoroughly tested. user: 'I just added a safety module that blocks…
Agent Claude Code
Use this agent when you need to build complex Rust CLI applications, especially those involving ML/AI integration, system-level programming, or cross-platform development. Examples: Context: User wants to create a sophisticated command-line tool with multiple backends and safety features. user: 'I need to build a Rust…
Agent Claude Code
Use this agent when you need to build command-line interface (CLI) applications in Rust, especially when implementing natural language processing tools, shell command generators, or any CLI that requires API integration, user interaction, and command execution. Examples: Context: User wants to create a CLI tool that…