helix-researchlisted

# Helix research loop Use this loop when a repository decision needs measured experiments rather than an implementation plan alone. ## Required launch protocol 1. Treat `$ARGUMENTS` as the exact research objective. If empty, ask and stop. 2. Read repository rules and define a named metric, its unit, a finite numeric target plus comparator (`lt`, `lte`, `eq`, `gte`, or `gt`), an exact measurement argv, and an exact test argv. Each argv must begin with a PATH-resolved executable name without `/`, `\\`, or a leading dash; use an interpreter for repository scripts. The measurement must emit exactly one JSON object shaped as `{ "metric": <name>, "value": <finite number>, "unit": <unit> }`. If that contract cannot be measured repeatably, stop. 3. Select `mode`: `original` by default, or explicit `graph`. Original uses `${CLAUDE_PLUGIN_ROOT}/workflows/helix-research.js`; graph uses `${CLAUDE_PLUGIN_ROOT}/workflows/graph/helix-research.js`. If a supplied mode is neither exact `original` nor exact `graph`, stop before doctor, evidence-session, or Workflow invocation. Configure `maxPasses` (`1..5`, default `5`), optional `plateauAfter` (`1..maxPasses`; omit to disable early plateau termination), and optional models for `planner`, `builder`, `tester`, `documenter`, `reviewer`, and `verifier`. 4. Run `node "${CLAUDE_PLUGIN_ROOT}/bin/helix-cc-doctor" --json`. Require a locally ready native or CLIProxyAPI workflow path. Stop for a global subagent-model override w