ptp-full

Run /ptp:full with a one-line change description or an existing openspec/changes/<id> folder. It decomposes the work into slices via plan-multiple, plan-reviews each slice, then applies and code-reviews them in order without requiring a second invocation.

/ptp:full is the union of the other two: full-plan decomposes and plan-reviews slices, full-apply applies and code-reviews them. /ptp:full runs both phases back to back, passing the planned slice ids explicitly so the apply phase skips its scope-confirmation stop.

Codex is required only when codex.mode is set to required; if it is missing the command stops immediately with an install-or-change-mode message. Under auto or off modes the run proceeds, and each phase skips the Codex reviewer non-silently while the Claude reviewer always runs.

Any slice ending in a non-green terminal state such as ITERATION CAP REACHED or ARTIFACT BUDGET EXCEEDED fails the plan-convergence gate, and the apply phase is never entered. With review.autoRecutOnBudgetExceeded enabled, budget-halted slices are automatically re-cut into smaller children subject to depth and growth caps.

Yes, when the parallel posture resolves to on and all four ptp-parallel-fanout safety conditions hold, one review-plan-full member is fanned out per slice, capped by parallel.maxConcurrency. After joining, a cross-slice dependency recheck runs before the convergence gate is applied.

Avoid it for small, single-unit changes where the full decomposition and dual-gate orchestration adds overhead; a plain /ptp:plan and apply flow fits better. It is also unsuitable when you want to inspect plans before any code is applied, since the handoff to the apply phase is automatic once planning converges.