{"skill":{"slug":"agent-arch","displayName":"Agent Arch","summary":"Agent loops, memory, tools, and safety boundaries. Use when designing AI agents.","description":"---\nname: agent-arch\ndescription: \"Agent loops, memory, tools, and safety boundaries. Use when designing AI agents.\"\n---\n\n# Agent Architecture Skill\n\nThis skill provides structured guidance for **Agent Architecture** work. Act as an active guide: confirm triggers, propose the stages below, and adapt if the user wants a lighter pass.\n\n## When to Offer This Workflow\n\n**Trigger conditions:**\n- User mentions **agent architecture** or closely related work\n- They want a structured workflow rather than ad-hoc tips\n- They are preparing a review, rollout, or stakeholder communication\n\n**Initial offer:**\nExplain the four stages briefly and ask whether to follow this workflow or work freeform. If they decline, continue in their preferred style.\n\n## Workflow Stages\n\n### Stage 1: Clarify context & goals\n\nAnchor on **goals, tools, and constraints**. Ask what success looks like, constraints, and what must not break. Capture unknowns early.\n\n### Stage 2: Design or plan the approach\n\nTranslate goals into a concrete plan around **memory and state**. Compare alternatives and explicit trade-offs; avoid implicit assumptions.\n\n### Stage 3: Implement, validate, and harden\n\nExecute with verification loops tied to **planning loops and stopping**. Prefer small steps, measurable checks, and rollback points where risk is high.\n\n### Stage 4: Operate, communicate, and iterate\n\nClose the loop with **safety monitoring**: monitoring, documentation, stakeholder updates, and lessons learned for the next cycle.\n\n## Checklist Before Completion\n\n- Goals and constraints are explicit for **Agent Architecture Skill**\n- Risks and trade-offs are stated, not hand-waved\n- Verification steps match the change’s impact (tests, canary, peer review)\n- Operational follow-through is covered (monitoring, docs, owners)\n\n## Tips for Effective Guidance\n\n- Be procedural: stage-by-stage, with clear exit criteria\n- Ask for missing context (environment, scale, deadlines) before prescribing\n- Prefer checklists and concrete examples over generic platitudes\n- If the user declines the workflow, switch to freeform help without lecturing\n\n## Handling Deviations\n\n- If the user wants to skip a stage: confirm and continue with what they need.\n- If context is missing: ask targeted questions before strong recommendations.\n- Prefer concrete examples, trade-offs, and verification steps over generic advice.\n\n## Quality Bar\n\n- Each recommendation should be **actionable** (what to do next).\n- Call out **failure modes** relevant to Agent Architecture (security, scale, UX, or ops).\n- Keep tone direct and respectful of the user’s time.\n","tags":{"latest":"1.0.0"},"stats":{"comments":0,"downloads":432,"installsAllTime":16,"installsCurrent":0,"stars":0,"versions":1},"createdAt":1774606995865,"updatedAt":1778492239922},"latestVersion":{"version":"1.0.0","createdAt":1774606995865,"changelog":"Initial release of the Agent Architecture Skill.\n\n- Provides a structured 4-stage workflow for agent architecture: clarify context & goals, design or plan approach, implement/validate, and operate/iterate.\n- Includes trigger conditions for when to propose the workflow.\n- Offers a checklist for thoroughness before marking work complete.\n- Emphasizes step-by-step guidance, risk transparency, and verification steps.\n- Adapts to user preferences (structured or freeform).\n- Supplies tips and quality standards focused on actionable advice and concrete examples.","license":"MIT-0"},"metadata":null,"owner":{"handle":"codenova58","userId":"s173fekm3yw84k7gp861dme7dd83gvyf","displayName":"codenova58","image":"https://avatars.githubusercontent.com/u/191358186?v=4"},"moderation":null}