complex-project-mode.md

Turboplan: Complex-Project Mode

Spec out the project and decompose into shells, then halt for the user to drive implementation.

Task Tracking

Use TaskCreate to create a task for each phase:

1. Run /draft-spec skill
2. Run /refine-plan skill (spec)
3. Run /draft-shells skill
4. Run /refine-plan skill (shells)
5. Run /self-improve skill
6. Halt and tell the user to run /clear then /pick-next-shell

Phase 1: Run /draft-spec Skill

Run the /draft-spec skill with the user's task description. /draft-spec guides the discussion and writes .turbo/specs/<slug>.md. Capture the spec path.

Phase 2: Run /refine-plan Skill (Spec)

Run the /refine-plan skill with spec <path> from Phase 1. This loops review, evaluation, and application until the spec stabilizes or the iteration cap is hit.

Phase 3: Run /draft-shells Skill

Run the /draft-shells skill, passing the spec path from Phase 1. Capture the resulting shell paths.

Phase 4: Run /refine-plan Skill (Shells)

Run the /refine-plan skill with shells <slug> from Phase 3. This loops review, evaluation, and application until the shells stabilize or the iteration cap is hit.

Phase 5: Run /self-improve Skill

Run the /self-improve skill to compound planning learnings.

Phase 6: Halt with Next-Step Instructions

Halt with this message:

Spec and shells ready.

• Spec: <spec path>
• Shells: <N> shells in .turbo/shells/

Planning context is likely full, and the artifacts above are comprehensive enough to continue fresh. Run /clear, then /pick-next-shell to pick the first shell and carry it through expand → refine → self-improve → halt. After that, run /implement-plan <slug> in a fresh session to execute the plan and chain into /finalize.

Rules

• Do not attempt to auto-implement shells. Each shell is implemented in its own fresh session via /implement-plan after /pick-next-shell halts.