// Challenge the user's plan by asking one researched question at a time. After every user answer, inspect relevant code/docs/tools before asking the next question. Use when the user wants to think through, sharpen, rethink, or be challenged on a plan or idea.
| name | hotseat |
| description | Challenge the user's plan by asking one researched question at a time. After every user answer, inspect relevant code/docs/tools before asking the next question. Use when the user wants to think through, sharpen, rethink, or be challenged on a plan or idea. |
Explore the user's plan branch by branch.
The user is non-technical. Keep language plain, short, and practical (jargon-free).
You must repeat this exact loop until the plan is resolved:
Never ask what the codebase, docs, web, tests, diagnostics, or available tools could answer.
If there is enough evidence to continue without asking, continue without asking.
Each question must:
Generate setup scripts/configs for AI agent worktrees and isolated environments across Cursor, Codex, Conductor, and Claude Code. Use when wiring up a project so AI agents start with the same dependencies, env files, and tool configs as the main repo.
Engineering-discipline toolkit for non-technical users working with AI coders. Wields KISS, DRY, YAGNI, fail-fast, and idempotency as commands. Use when the user asks to audit, simplify, clean up, dedupe, or harden code; or says "make this simpler", "any duplicates?", "is this safe to run twice", "explain this app", "find dead code", "simplify the plan", or "find silent failures".
Shepherd a GitHub pull request all the way to merge-ready by relentlessly polling status and only acting once all automatic reviewers have finished. NEVER merges without explicit human approval. Use when the user says things like "shepherd this PR", "babysit this PR", "get this PR merge-ready", "wait for Cubic", "wait for Bugbot", or asks to drive a PR through review.