{"slug": "show-hn-agent-architect-skill-for-building-agentic-systems", "title": "Show HN: Agent Architect Skill for building agentic systems", "summary": "A developer released Agent Architect, a skill for Claude Code and Codex that evaluates and designs multi-agent AI systems, offering AUDIT, REVIEW, and DESIGN modes. The tool scores systems across 11 dimensions including prompt architecture, tool design, context management, and production readiness, and installs into both ~/.claude/skills and ~/.agents/skills via a setup script.", "body_md": "Claude Code and Codex skills for evaluating and designing multi-agent systems.\n\nSenior architect review for agent systems, prompt engineering, and harness design.\n\n**Three modes:**\n\n**AUDIT**— Full 11-dimension evaluation of an existing agent system. Scores prompt architecture, tool design, context management, multi-agent orchestration, eval infrastructure, production readiness, model awareness, agent security, memory architecture, harness architecture, and multimodal architecture. Produces a maturity score with prioritized recommendations.**REVIEW**— Focused teardown of a specific prompt, skill file, or tool definition. Line-by-line findings with confidence scores.** DESIGN**— Architect a new agent system from scratch. Produces system prompt drafts, tool specs, failure mode maps, eval plans, and implementation checklists.\n\nThe current agent-architect version is built for production AI systems, not just prompt review. It evaluates runtime API contracts, tool harness boundaries, MCP/tool governance, background task execution, trace-based evaluation, and voice/image/video surfaces when those modalities are present.\n\n`./setup`\n\ninstalls each skill into both `~/.claude/skills`\n\nand `~/.agents/skills`\n\n.\n\n```\ngit clone git@github.com:YOUR_ORG/agent-skills.git ~/.claude/skills/agent-skills\ncd ~/.claude/skills/agent-skills\n./setup\n```\n\nOr install from any location:\n\n```\ngit clone git@github.com:YOUR_ORG/agent-skills.git ~/path/to/agent-skills\ncd ~/path/to/agent-skills\n./setup\n```\n\nIn Claude Code:\n\n```\n/agent-architect\n```\n\nIn Codex, once installed under `~/.agents/skills`\n\n, reference `agent-architect`\n\nby name in your request.\n\nThe skill auto-detects your codebase and selects the appropriate mode. You can also be explicit:\n\n```\n/agent-architect audit this system\n/agent-architect review this prompt: path/to/prompt.md\n/agent-architect design a new agent for customer support\n```\n\nCreate a new directory with a `SKILL.md`\n\nfile:\n\n```\nagent-skills/\n├── agent-architect/SKILL.md    # existing\n├── my-new-skill/SKILL.md       # new skill\n└── setup                       # auto-discovers all */SKILL.md\n```\n\nRun `./setup`\n\nto register the new skill.", "url": "https://wpnews.pro/news/show-hn-agent-architect-skill-for-building-agentic-systems", "canonical_source": "https://github.com/anyaa-labs/agent-skills", "published_at": "2026-08-01 12:50:23+00:00", "updated_at": "2026-08-01 13:22:28.394698+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "developer-tools", "ai-infrastructure"], "entities": ["Agent Architect", "Claude Code", "Codex"], "alternates": {"html": "https://wpnews.pro/news/show-hn-agent-architect-skill-for-building-agentic-systems", "markdown": "https://wpnews.pro/news/show-hn-agent-architect-skill-for-building-agentic-systems.md", "text": "https://wpnews.pro/news/show-hn-agent-architect-skill-for-building-agentic-systems.txt", "jsonld": "https://wpnews.pro/news/show-hn-agent-architect-skill-for-building-agentic-systems.jsonld"}}