| Fix a bug / implement a feature in a repo | anthropic-agent-sdk + sdk | tools, sandbox(edit), budget | Minimal single agent |
| Same, but survive outages/rate limits | + ordered fallback to openai-codex-sdk, gemini-antigravity-cli | selection, retry | Resilient worker |
| Research a topic, then build, then review | gemini-genai-sdk → anthropic-agent-sdk → openai-codex-sdk | pair_profiles | Pair recipe |
| Get an unbiased code review | openai-codex-sdk + cli, isolate_context: true | tool_policy(read-only) | Pair recipe |
| Gate completion on multi-vendor agreement | any worker + review panel | review_consensus | Review panel |
| Extract structured data (JSON) | openai-responses-api + api or any SDK runtime | state_machine + output_schema | Structured Output |
| Cheap bulk classification/evals | batch surfaces | stage that submits, stage that collects | Cost, Batch & Cache |
| Multi-stage pipeline with typed handoffs | any | state_machine | State machine |
| Hard problem — buy more reasoning | Anthropic thinking / OpenAI reasoning_effort / Antigravity level | agent.thinking | Thinking |
| Do a web task with no API (forms, dashboards) | openai-operator + api | budget, approval-friendly policy | Computer Use |
| Voice panel of experts | roster with voice/voice_name | roster | Voice & Realtime |
| Keep data on-device / offline | local-llm + api | budget | Local provider |
| Block specific tool calls deterministically | Claude-native hooks (settings) | tool_policy as backstop | Hooks |
| Plug in external SaaS tools | participant with mcp_server_config | tools | Tools & MCP |
| Cut spend on repeated context | — | cache_policy, Headroom | Cost, Batch & Cache |