What problem does it solve? Running the phase-loop orchestration runtime from OpenCode requires knowing the correct CLI commands, shared state files, and blocker semantics; this Skill gives OpenCode operators a thin bridge to the repo-local runner without reimplementing phase selection, reconciliation, or execution logic. ## Core Features & Use Cases - Command Mapping: Translates OpenCode intents (handoff, status, monitor, resume, run, dry-run, sync-skills) into the correct phase-loop or codex-phase-loop CLI invocations with bounded defaults like --max-phases 1. - Shared State Inspection: Reads .phase-loop/state.json, events.jsonl, handoff files, and run heartbeats to report automation status, verification status, and frozen blocker literals such as branch_sync_conflict. - Drift and Blocker Diagnosis: Detects installed-skill drift (missing_root, missing_skill, drifted) and guides repair via sync-skills --apply, plus branch-governance preflight handling for pipeline-mode dispatches. - Use Case: An OpenCode user asks to resume a stalled phase-loop run; the Skill checks monitor --once --json for the current blocker, confirms the worktree state, and issues a bounded phase-loop run --max-phases 1 --closeout-mode manual. ## Quick Start Ask the assistant to check phase-loop status from OpenCode, for example: "Show me the current phase-loop status and any blockers before I resume the run."
