{"skill":{"slug":"agents","displayName":"Agents","summary":"Design, build, and deploy AI agents with architecture patterns, framework selection, memory systems, and production safety.","description":"---\nname: Agents\ndescription: Design, build, and deploy AI agents with architecture patterns, framework selection, memory systems, and production safety.\n---\n\n## When to Use\n\nUse when designing agent systems, choosing frameworks, implementing memory/tools, specifying agent behavior for teams, or reviewing agent security.\n\n## Quick Reference\n\n| Topic | File |\n|-------|------|\n| Architecture patterns & memory | `architecture.md` |\n| Framework comparison | `frameworks.md` |\n| Use cases by role | `use-cases.md` |\n| Implementation patterns & code | `implementation.md` |\n| Security boundaries & risks | `security.md` |\n| Evaluation & debugging | `evaluation.md` |\n\n## Before Building — Decision Checklist\n\n- [ ] **Single purpose defined?** If you can't say it in one sentence, split into multiple agents\n- [ ] **User identified?** Internal team, end customer, or another system?\n- [ ] **Interaction modality?** Chat, voice, API, scheduled tasks?\n- [ ] **Single vs multi-agent?** Start simple — only add agents when roles genuinely differ\n- [ ] **Memory strategy?** What persists within session vs across sessions vs forever?\n- [ ] **Tool access tiers?** Which actions are read-only vs write vs destructive?\n- [ ] **Escalation rules?** When MUST a human step in?\n- [ ] **Cost ceiling?** Budget per task, per user, per month?\n\n## Critical Rules\n\n1. **Start with one agent** — Multi-agent adds coordination overhead. Prove single-agent insufficient first.\n2. **Define escalation triggers** — Angry users, legal mentions, confidence drops, repeated failures → human\n3. **Separate read from write tools** — Read tools need less approval than write tools\n4. **Log everything** — Tool calls, decisions, user interactions. You'll need the audit trail.\n5. **Test adversarially** — Assume users will try to break or manipulate the agent\n6. **Budget by task type** — Use cheaper models for simple tasks, expensive for complex\n\n## The Agent Loop (Mental Model)\n\n```\nOBSERVE → THINK → ACT → OBSERVE → ...\n```\n\nEvery agent is this loop. The differences are:\n- What it observes (context window, memory, tool results)\n- How it thinks (direct, chain-of-thought, planning)\n- What it can act on (tools, APIs, communication channels)\n","tags":{"latest":"1.0.0"},"stats":{"comments":0,"downloads":2074,"installsAllTime":78,"installsCurrent":13,"stars":3,"versions":1},"createdAt":1770983222662,"updatedAt":1778489510625},"latestVersion":{"version":"1.0.0","createdAt":1770983222662,"changelog":"Initial release: architecture patterns, frameworks comparison, use cases, implementation, security, evaluation","license":null},"metadata":null,"owner":{"handle":"ivangdavila","userId":"s178jdk12x4qj3gs2se3etxf3h83h7ft","displayName":"Iván","image":"https://avatars.githubusercontent.com/u/81719670?v=4"},"moderation":null}