daily-startlisted

Session-start routine: detect mode, load VP context, show routines + pending, then either prompt the user for goals (human) or chain to `dispatch-task-start` on the highest-priority unblocked task (autonomous). When the user's stated goal needs a new mission scaffold, chain to `mission-bootstrap` instead of raw `create_mission`. **Canonical source**: VantageRegistry (`get_skill_content name=daily-start`). The local `.claude/skills/daily-start/SKILL.md` in each workspace MUST be a byte-exact mirror of the VR canonical content. End of hand-copy — fetch from VR, do not edit locally. V3 PRINCIPLE — NEVER RAW `start_task` AT SESSION-START. Every transition from "task picked" to "task running" goes through `dispatch-task-start` so the IRP-sequence hook (`enforce-irp-sequence.py`) cannot block on a stale `in_progress` task left over from the previous session. Same rule applies whether Pi confirms a human-mode pick or sigma/eta auto-pick autonomously. V3 PRINCIPLE — NEW MISSION ⇒ `mission-bootstrap`. When the user's stated goal is "start a mission for X" / "scaffold mission Y" / "new mission Z", do not call `create_mission` directly. Chain to the `mission-bootstrap` skill, which wires VERIFICATION/TESTS-compliant child tasks and pre-satisfies the task-quality hook. ## WORKFLOW **Step 1 — Detect mode (human vs autonomous)** HUMAN MODE is opt-in via a HOST-SIDE marker that lives outside every repo and every published package. Never key it on a hardcoded filesystem path or a perso