superpowers

ey still do make mistakes. In fact, they are now at a level where the hallucinations aren't obvious making it even more important to keep a close eye on the result. If you do want to lean more heavily into agents doing most of the work, try to make sure they are following proper development practices. Something they don't actually do by default but using something like the superpowers skills makes a world of difference: https://github.com/obra/superpowers Having them follow TDD helps a lot. I've even considered adding a QA agent/skill in here expanding things further to not just unit tests and some basic manual tests but also creating proper automated tests (following the test automation pyramid principles) to create an entire test suite. Not to actually give more control to LLMs, but because it allows me to more easily spot where things go sideways. Since more tests, including e2e tests and UI tests where possible let me review more aspects of the work they do. Having said that, I haven't created that skill yet. As reviewing the work of multiple agents is already exhausting as is. I am not in a position where I have to use AI or else in my company so instead I have dialed down my agentic usage by quite a lot to the point where I barely use agents anymore. To me using LLMs mostly as tools outside the process still is the sweet spot.

ut organisational management they can't collaborate effectively. It's like having a room full of talented people with no management structure — capability is there, coordination isn't. Human companies solved this centuries ago. Performance reviews, SOPs, org hierarchy, hiring pipelines, retrospectives — these mechanisms make collaboration work regardless of domain. We applied the same principles to AI agent teams. *How it works* The company has a founding team (HR, COO, EA, CSO) built in from day one. Talent Market. When the company needs specialists, HR recruits from a community-driven Talent Market ([one-man-company.com](https://one-man-company.com)). Verified agent implementations with real usage counts and community reviews, not LLMs self-declaring expertise. Claude Code, OpenClaw, LangGraph agent families are already on the market — [superpowers](https://github.com/obra/superpowers), [agency-agents](https://github.com/msitarzewski/agency-agents) are ready to hire ([browse here](https://one-man-company.com/talent/claude-superpowers)). The market is open for contributions: upload your own agents, share/buy/sell. We're building a real agent community. Talent-Container separation. Each agent is split into a portable identity (Talent: prompts, skills, tools, work principles) and an execution runtime (Container: LangGraph, Claude CLI, script process). Different agents run on completely different backends in the same project — Claude Code writes code, LangGraph handles image generation, OpenClaw runs admin. Tree Search based Orchestrate Algorithm. Tasks decompose dynamically during execution, not upfront in a fixed pipeline. The underlying algorithm is built on AND-tree DAG structure with formal guarantees: termination (bounded retries + timeouts + cost budgets), deadlock freedom (cycle detection at insertion), crash recovery (processing nodes reset on restart). Quality gates. Every task has acceptance criteria and mandatory supervisor review. Self-evolution. Post-project retrospectives update work principles and org-wide SOPs. CEO 1-on-1s become permanent agent memory. Performance reviews with real consequences: PIP for underperformers, then offboarding if they don't improve. HR re-recruits from the Talent Market. *Results* Cross-domain without changing the framework. So far, we have tried 1) Game dev; 2) Audiobook and Short Movie Generating; 3) Auto-Researchurvey; 4) Content Generation; 5) More is comming! Same system, different team, different domain. Build a game today, produce a comic series tomorrow. *Try it* Fully open source. [one-man-company.com](https://one-man-company.com) is the front door. One command: GitHub: https://github.com/1mancompany/OneManCompany Talent Market: https://one-man-company.com Come try it, file issues, or contribute. We especially want people building agents for the Talent Market.

reference the skill by name. [1]: https://github.com/obra/superpowers

er and safer to read, understand, refactor, modify, extend, test, and maintain. Arxitect helps coding agents build robust software through best practice design and architecture. It helps coding agents write better code. And in turn, coding agents become more efficient and more accurate. I was inspired by the design of [Superpowers](https://github.com/obra/superpowers) and extended that concept to software design principles: * API Design - Assessing naming conventions, method signatures, parameter design, type safety, and REST endpoint design. * Object Oriented Design - Assessing SOLID principles, DRY violations, composition vs. inheritance choices, and design pattern applicability. * Clean Architecture - Assessing component cohesion, component coupling, and quality attributes like maintainability, extensibility, and testability. Install in Claude Code with: /plugin marketplace add andonimichael/arxitect /plugin install arxitect@arxitect /reload-plugins I welcome any feedback, advice, or PRs at: https://github.com/andonimichael/arxitect