Autoconsultant

Recipe

1. Run Consultant to an approved or explicitly accepted plan.md, preserving its explicit references and implementation boundary.
2. Keep one session-wide Consultant Question <Q> sequence across both phases, never invoking AskUserQuestion and rendering every inherited discovery, review, blocker, artifact approval, and revision prompt as a clear bold question, necessary context below it, then numbered options with option 1 ending (Recommended) and final Other.
3. Present the Optimizer approval question with the complete instruction as its context, using the approved plan and inspected repository as Automake's source of truth and excluding evaluation criteria.
4. Present a separate Evaluator approval question with the complete independent instruction as its context, excluding Optimizer rationale and prior evaluation material.
5. Resolve maximum iterations, consecutive-failure limit, and observable success condition from supplied values or Automake defaults.
6. Emit one dynamically safe fenced handoff containing absolute repository and plan paths, complete approved role instructions, every resolved value, the plan-preservation contract, and directions for a fresh agent to invoke Automake run-now as already approved without repeating setup.
7. Stop after the handoff without showing Automake's Run gate, performing Git preflight, mutating the repository, creating Automake state, launching a child, running the ratchet, or implementing.

Details

• Keep every scratch artifact and fence check inside the declared run directory or in memory; never use an external temporary directory.
• Direct the fresh agent to read plan.md and every explicit reference it names before invoking Automake.