ptp-full-apply

Run /ptp:full-apply with explicit change ids, an epic: or story: selector, or no arguments to discover active changes via openspec list. The workflow applies each story fully, then runs its review, and only moves to the next story after the review converges.

The workflow halts the entire run at that story rather than continuing. The terminal report buckets stories into processed, applied (review pending), and never-started, and prints a resume command for the unprocessed tail.

Only when codex.mode resolves to required; then the command runs codex --version and stops without launching if Codex is missing. Under auto or off it launches normally, and each story's review skips the Codex phase non-silently as gate-success.

The command reads each story's effort.md and parses line 1 as model.effort, deriving the review target by flooring the model at sonnet and effort at high. A missing or unparseable file defaults to opus.high and is noted, never crashing the run.

Yes, exactly once per story. If a review's fix target names a more capable model, the workflow re-spawns that review at the escalated target and reads its terminal state; an escalation that cannot be honored halts the story with an explicit reason.

With no ids it discovers all active changes, which may include unrelated or unplanned work, so it prints the resolved ordered list and stops before any apply. Re-invoking with the confirmed set or explicit ids proceeds without the gate.