odin-run-advanced

Use when asked to run /artifact-arena to generate and judge competing artifact implementations. Produce one verified synthesis with provenance and rejection reasons. Don't use for remote, credential, publish, deploy, or irreversible changes.

Use when a human has an approved delivery plan and wants the full chain run under phase gates. Delegates implementation to work, simplification to simplify, review to review, and finalization to review-and-ship. Not for planning, open-ended debugging, or single-step execution — use work directly.

Use when a human explicitly runs /cloud-task-orchestrator for a large task across cloud agents to drain a verified task graph. Not for local subagent coordination — use orchestration-patterns.

Use when a candidate patch or answer needs independent corroboration before it is trusted. Run N isolated sandboxed attempts and return the strict-majority byte-identical patch with its agreement count, or report no consensus. Not for multi-stance investigation — use council.

Use when a configured loop misbehaves, produces unexpected results, or its setup soundness is questioned. Reads the charter, state, gate, and budget files, classifies health as healthy, warning, or blocked, and emits at most three prioritized actions. Not for tasks that require source or remote-system changes; not for loop design — use harness-engineering.

Use when asked to /feedback-sweep [setup|reconfigure] [mode:non-interactive]: sweep configured sources, acknowledge user-approved items, verify fixes merged to default, and emit an autopilot-ready plan. Not for pattern discovery — use feedback-pattern-sweep.

Use when asked to design a bespoke execution workflow when no playbook fits; verify a falsifiable predicate on the real product with a reviewable trail. Don't use for remote, credential, publish, deploy, or irreversible changes.

Use when asked to evaluate a proposed commit, merge, or auto-merge against a version-1 gate config. Returns an allow or deny verdict naming matched paths; performs no merge. Not for landing PRs — use gate-and-merge. Read-only.

Use when a human runs /merge-and-deploy to merge a PR and trigger or verify deployment. Don't use for tasks that require source changes or without explicit human confirmation at each gate.

Use when a high-priority run reaches at least 90% of its budget and requests an extension. Ask its human owner once; on grant record the bounded extension, on decline mark the run WAITING_FOR_BUDGET. Never self-grants. Not for low-priority runs — let those exhaust silently.

Use when work decomposes across subagents or role panels: select a coupling-based orchestration mechanism, spawn bounded subagents, verify every artifact owner-side, and write only the reconciled synthesis. Not for cloud-agent task-graph orchestration — use cloud-task-orchestrator.

Use when work splits into independent sub-tasks or cross-domain research. Runs concurrent agents and returns one reviewed composition or an exact gap. Not for durable scope ownership with per-scope review — use partition-scopes-to-subagents.

Use when asked to partition non-overlapping scopes across subagents and drive each to completion with fresh review, fix reports, and a glanceable actual-work map. Not for concurrent independent sub-tasks — use parallel-launch.

Use when asked to apply the pstack rigor mode to non-trivial work. Routes the task to a matched playbook from a 23-index and executes it to verified completion on the real surface. Not for remote, credential, publish, deploy, or irreversible changes.

Use when asked to change state in an external system, propose the write locally and halt at the human gate without executing it. Don't use for remote, credential, publish, deploy, or irreversible changes.

Use when a human runs /publish-release-pr to ship a versioned, redaction-gated PR with evidence. Not for tag-and-push releases — use release-gate.

Use when a user runs saga or asks to autonomously build a sizable feature. Produces a spec tree, delegates to worker subagents in isolated worktrees, validates at each milestone, and gates user acceptance before VCS commit. Not for executing a given plan — use subagent-driven.

Use when a user says execute with subagents or hands over an ordered multi-task plan to run a per-task implementer-reviewer loop with an audit gate. Don't use for credential, paid, or deployment work, or for any mutation outside the skill's designated workspace.

Use when asked to run partitioned parallel coverage or races across isolated workers, producing a consolidated evidence table with gaps and dropouts. Don't use for remote, credential, publish, deploy, or irreversible changes.

Use when a scheduled or watcher tick fires and a lightweight pulse must capture current external state. Not for remote, credential, publish, deploy, or irreversible changes.

Use when asked to drive decomposable work as a deterministic multi-subagent workflow with phased fan-out and adversarial verification. Also handles audits, migrations, broad research sweeps, or scale one context cannot hold. Don't use for remote, credential, publish, deploy, or irreversible changes.