Propose Ship Auto

Run the full minerva lifecycle end-to-end with consensus-panel decisions in place of human gates. This skill is a hybrid orchestrator — it delegates to minerva:ship and minerva:cleanup directly (those phases have no strategic gates) but inlines the propose / work / review / promote / replan phases so it can substitute panel calls for hard user gates.

The mechanism: at each strategic or tactical decision point, dispatch a 3-agent Proponent/Skeptic/Arbiter panel of fresh-context subagents — the panel mechanics live in minerva:round-table, to which this skill delegates (see the Delegation section of references/panel-protocol.md). Operational decisions (commit messages, PR bodies, file paths) bypass the panel entirely — the main LLM executes them. For small, low-risk decisions, a Skip predicate (references/panel-protocol.md) lets the main LLM decide directly without convening a panel — so a genuinely small task runs effectively panel-free — while it fails closed to the full panel on any uncertainty and never skips the post-divergence or completion-verification panels.

Usage

• minerva:propose-ship-auto "add rate limiting" — start a new auto run with the inline description as the strategic seed.
• minerva:propose-ship-auto — start with current-session chat context as the strategic seed (only sensible if the chat already discussed what to build).
• minerva:propose-ship-auto --cleanup-only <date-slug> --retry=N — internal re-entry from the cleanup-gate wake-up loop. Skips phases 1–6 and re-runs Phase 7.

Pre-flight: in-flight work collision

Identical to minerva:propose-ship's pre-flight section. This check is not panel-decided — wrong call here destroys real work, so escalation to the user is hardcoded.

Read plugins/minerva/skills/propose/references/in-flight-check.md and run it. It reads four evidence sources — local work units (via the in_flight predicate, never a string match), local and remote branches, open PRs, and live sibling Claude sessions — each failing soft, so a repo with no remote, no tracker and no siblings passes through silently. It is detection, not a lock: git worktree add -b serializes only sessions choosing the same slug, so a clean result means no evidence was found, not that nobody else is working the goal.

When a peer session messages you, read plugins/minerva/skills/propose/references/cross-session.md: inform, never delegate.

A collision is a hardcoded AskUserQuestion (resume that work / start fresh anyway / abandon this run) and increments the global escalation counter.

Only proceed after the user confirms. This is the single mandatory — and only permitted — pre-run user interaction (see No ceremony ratification in references/panel-protocol.md).

Panel protocol

The full panel-protocol policy — the Skip predicate (small decisions), No ceremony ratification, Delegation to minerva:round-table, and Per-decision logging (formats and worked examples) — lives in references/panel-protocol.md. Read that file once, in full, before this run's first strategic/tactical decision point; its rules then apply to every decision that follows.

Binding floor, even before the reference is read:

• The conjunctive skip predicate is applied silently per-decision, only where the Decision taxonomy row (in references/panel-protocol.md) is skippable, and fails closed to a panel on any uncertainty. It is the only de-ceremony mechanism — never ask the user to pre-ratify skips or pick a "ceremony level".
• Never-skippable: completion verification, mid-work divergence confirmation, new-plan acceptance (replan), Replan-vs-FIX, and every hardcoded escalation trigger.
• Panel mechanics (dispatch, Proponent/Skeptic/Arbiter briefs, votes, revision round, escalation composition) are delegated to minerva:round-table in caller mode; quorums always come from the Decision taxonomy (in references/panel-protocol.md), never round-table's standalone default.
• Every panel call, predicate skip ([skipped — small] + evidence), and user-directed bypass ([user-directed]) logs one line to scratchpad.md under ## Panel decisions YYYY-MM-DD.

Phases

Execute the phases in order. The full inline protocols — panel artifacts, vote handling, escalation aftermath, and file-write steps — live in references/phases.md. Before executing each phase, read that phase's section there; the map below locates the work, it is not the protocol:

1. Propose (inline) — assemble context → design synthesis → scope-check panel (3/3) → approach-selection panel (3/3) → whole-proposal-acceptance panel (3/3) → worktree + branch + file writes per minerva:propose → self-review. No post-write user gate.
2. Work (inline) — implement per minerva:work's protocol; divergence panel (2/3) when a load-bearing divergence is suspected; completion-verification panel (3/3) on the success-criteria checklist + diff.
   • 2.5 Replan (inline, if triggered) — draft Original plan / What changed / New plan; new-plan-acceptance panel (3/3); append to replan.md.
3. Review (inline) — minerva audit + code review (PR mode delegates to code-review:code-review); single triage panel (2/3) over all findings; replan-vs-FIX panel (2/3) if a load-bearing finding surfaces.
4. Promote (inline) — partition panel (2/3); TODO-disposition panel (2/3); apply writes per minerva:promote Mode A; archive scratchpad.
5. Ship gate — no gate: silent advancement, except halt if the global escalation counter has reached 3.
6. Ship (delegated) — invoke minerva:ship via the Skill tool with its auto-mode instruction (auto-accept hard gates #1 commit message and #2 PR title/body; everything else unchanged). CI auto-fix bails classified other are escalated to the user — never panel-voted.
7. Cleanup gate — poll PR state via gh pr view; on MERGED invoke minerva:cleanup via the Skill tool with args <date-slug> --yes (which also reconciles the knowledge wiki and opens its auto-merging PR); on OPEN with auto-merge, ScheduleWakeup re-entry (--cleanup-only <date-slug> --retry=N, cap 12); otherwise surface manual instructions.

Failure modes, escalation, budget caps

Binding caps: one initial vote + one revision vote per decision (6 subagent dispatches max); propose-phase abort when 2 of its 3 panels escalate; global escalation counter halts the run at 3. Hard escalation triggers that skip the panel entirely: in-flight collision, worktree-creation failure, ship-phase failures (other classification, push rejection, gh auth failure), counter at 3. The full trigger list, final-report-on-bail format, and observability requirements live in references/governance.md — read it at the first escalation or before reporting any bail.

Out of scope

Never modify any existing minerva skill at run time (this skill orchestrates by invocation only); never auto-cascade into new work units; never cap implementation time; review/promote ordering is fixed; quorums are not configurable. Rationale and detail: references/governance.md.