using-the-think-steplisted

Decide where an agent should reason during a task and how to prompt for it. Covers three places a model can think and how to choose between them: extended thinking before the turn, interleaved thinking between tool calls, and the no-op "think" tool that logs reasoning mid-chain at a point you pick. Also covers when a mid-task step helps, when it only adds latency, the contract interleaved thinking imposes on the request, and why the think tool's gains come from system-prompt guidance rather than the tool itself. Use this whenever someone builds an agent that follows policies or makes long sequential decisions, asks about a think or scratchpad tool or interleaved thinking, finds an agent skipping rules or mishandling tool output mid-task, or wants more deliberate tool use. Trigger on "think tool," "interleaved thinking," "reasoning between tool calls," "let the agent reason before acting," "agent ignores the policy," and similar. Not for general prompt engineering or one-shot chain-of-thought.
Hoja-Solutions/agent-stdlib · ★ 1 · AI & Automation · score 72

Install: claude install-skill Hoja-Solutions/agent-stdlib

# Using the think step Source: [The "think" tool](https://www.anthropic.com/engineering/claude-think-tool), with the cross-mode decision drawing on the [extended thinking docs](https://platform.claude.com/docs/en/build-with-claude/extended-thinking). The pattern ships as MCP servers around the web; none package the judgment of *when* to use it, how it relates to extended and interleaved thinking, and how to prompt it. This pack ships the tool itself under `mcp-servers/think/`; this skill is about reaching for it well. The think tool is a no-op the model can call to reason in the middle of a tool-use chain. It changes no state and returns nothing useful; its only job is to give the model a moment to lay out its reasoning before the next action. That makes it different from extended thinking, which happens once before the agent acts. The think step happens *between* actions, after the agent has seen a tool result and before it commits to what comes next. ## Where the reasoning goes A model can reason in three places, and each does a different job: - **Extended thinking, before the turn.** The model plans before it acts or fires off parallel, independent tool calls. On recent models, turn on adaptive thinking and set an effort level; on older ones, set a token budget below `max_tokens`. Reach for it when the hard part is the plan you make up front. - **Interleaved thinking, between tool calls.** The model reasons after each tool result, inside the same turn, before it picks