opensearch-review-workflowlisted

# OpenSearch Review Workflow Run the three declared OpenSearch review skills against one source snapshot. Validate their availability and complete execution independently of other installed OpenSearch skills. ## Declared inventory and dependencies Runnable skills, exactly: - `opensearch-optimization` - `opensearch-query-review` - `opensearch-schema-review` Optional synthesis agent: `opensearch-expert`. `opensearch-review-workflow` is the orchestrator and is never recursively invoked as a component. ## Cost-aware native routing Resolve portable profiles from the current runtime capability catalog at dispatch. Use the three mandatory skill passes at `frugal/low` for a bounded static diff, promote only the affected query, schema, or optimization pass to `standard/medium` for cross-index or rollout reasoning, and use the optional `opensearch-expert=standard/medium` only when specialist evidence conflicts. Reserve `frontier/high` for confirmed irreversible data or public compatibility risk. Every native handoff uses `model_selection: explicit-native-override`; spawn with `fork_turns: none`, the exact resolved `model`, and exact `reasoning_effort`. Never inherit the primary model or reasoning setting. Persist requested/observed routes in the ledger and reject mismatch. Timeout, quota, authentication, and missing cluster access do not justify a larger model. ## Input contract Accept a repository/diff scope and any supplied mappings, templates, settings, queries, pipelines