/effort

Choose how hard the model thinks before answering

• Instructions: # /effort — how hard the model thinks Reasoning models spend time THINKING before they answer, and how much is a per-request setting (`reasoning_effort` on the OpenAI wire, which OpenRouter forwards to the model). This skill picks that setting for the next messages in this chat. The value comes from the first of these that says something: 1. **Your choice here** — stored on your composer, so it follows you across tabs and threads until you change it. It wins over everything below. 2. **The agent's own `reasoningEffort`** — every shipped agent declares one (chat and utility agents `medium`, coding agents `xhigh`). 3. **The model node's own `reasoningEffort`** — what an admin set on the `LanguageModel` node for everyone. 4. **The tier's default** — the agent's tier, or the `chat` tier when the agent names none (`chat` and `utility` are `medium`, `coding` is `xhigh`). Choose *Model default* above to fall back to rungs 2–4. The provider's own default is reached only when all of these are blank — and for a reasoning model it is usually its SLOWEST setting: a model whose reasoning is on at maximum by default can spend a minute thinking before each tool call, so a round of nine cheap tool calls takes minutes instead of seconds. Turning the effort down is the fix, and it costs nothing on models that have no reasoning phase — they ignore the field. Under the Claude Code harness `/effort` belongs to the CLI instead and is forwarded to it 1:1; the levels are the CLI's own.
• Action › Kind: 0
• Action › Field: effort
• Action › Title: Choose reasoning effort
• Auto mount: yes
• Launches sub thread: no