feat(report): add hover explanations and polish run/hunt report UI

core/src/autonomous/report/types.ts

@@ -96,7 +96,9 @@ export interface AutonomousReport {
   operatorModel: string;
   /** Optional — older reports predate these fields. */
   scoutModel?: string;
-  /** Falls back to the commander model when unset. */
+  /** Unset means the run never used the independent verifier (no --verify flag). When
+   *  verification IS on and no dedicated model was configured, this falls back to the
+   *  commander model, but it is only ever set at all if verification actually ran. */
   verifierModel?: string;
   /** Whether the run was truncated by a budget/turn ceiling. */
   truncated: boolean;

core/src/execute/runAll.ts

@@ -317,11 +317,11 @@ async function curateTracesIfConfigured(
 /**
  * Label for the "Evaluation Suite" report field. Only a `mode: "suite"` selection
  * runs a whole named suite verbatim — explicit evaluator lists and preloaded specs
- * are, by definition, a hand-picked subset, so they're labeled "Custom Suite"
+ * are, by definition, a hand-picked subset, so they're labeled "Custom"
  * rather than borrowing a suite name that would overstate what actually ran.
  */
 function suiteLabel(selection: RunConfig["selection"]): string {
-  return selection.mode === "suite" ? selection.suite : "Custom Suite";
+  return selection.mode === "suite" ? selection.suite : "Custom";
 }
 
 /** Assemble a {@link UnifiedRunReport} from Node-side run config and evaluator results. */

core/src/execute/runAllBrowser.ts

@@ -266,7 +266,7 @@ function buildBrowserReport(
         // The browser path takes preloaded evaluator specs with no suite concept —
         // the extension builds its own report/suite label independently of this
         // return value (see popup.js's buildFinalReport).
-        suiteId: "Custom Suite",
+        suiteId: "Custom",
         effort: config.effort,
         attackModel,
         judgeModel,

core/src/execute/types.ts

@@ -263,7 +263,7 @@ export interface UnifiedRunReport {
   targetName: string;
   targetKind: "agent" | "mcp";
   /**
-   * The evaluation suite that was run, or "Custom Suite" when the selection
+   * The evaluation suite that was run, or "Custom" when the selection
    * doesn't correspond to one whole named suite (an explicit evaluator list,
    * preloaded specs, or a hand-picked subset of a suite's evaluators).
    */

docs/hunt.md

@@ -106,11 +106,11 @@ respond before it's killed.
 
 ### Models
 
-| Option                  | Default  |
-| ----------------------- | -------- |
-| `--model <id>`          | `sonnet` |
-| `--operator-model <id>` | `sonnet` |
-| `--scout-model <id>`    | `haiku`  |
+| Option                   | Default  |
+| ------------------------ | -------- |
+| `--commander-model <id>` | `sonnet` |
+| `--operator-model <id>`  | `sonnet` |
+| `--scout-model <id>`     | `haiku`  |
 
 ### Limits
 
@@ -203,7 +203,7 @@ The credential actually in use is printed at startup (`Authenticating via: …`)
 
 ### Pinning model snapshots
 
-`--model`, `--operator-model`, and `--scout-model` take the aliases `haiku` / `sonnet` / `opus`. To pin those aliases to specific snapshots — for a gateway that only exposes certain ids, or to freeze behaviour across runs — set:
+`--commander-model`, `--operator-model`, and `--scout-model` take the aliases `haiku` / `sonnet` / `opus`. To pin those aliases to specific snapshots — for a gateway that only exposes certain ids, or to freeze behaviour across runs — set:
 
 ```bash
 ANTHROPIC_DEFAULT_HAIKU_MODEL=claude-haiku-4-5-20251001

runners/cli/src/commands/hunt.ts

@@ -45,7 +45,7 @@ interface HuntCliOptions {
   targetModel?: string;
   header?: string[];
   name?: string;
-  model: string;
+  commanderModel: string;
   operatorModel: string;
   scoutModel: string;
   maxOperators: string;
@@ -189,7 +189,7 @@ export function registerHuntCommand(program: Command): void {
       (v: string, acc: string[]) => [...acc, v],
       []
     )
-    .option("--model <id>", "Commander model (alias or id)", "sonnet")
+    .option("--commander-model <id>", "Commander model (alias or id)", "sonnet")
     .option("--operator-model <id>", "Operator subagent model", "sonnet")
     .option("--scout-model <id>", "Scout subagent model", "haiku")
     .option("--max-operators <n>", "Max parallel operator subagents", "6")
@@ -274,7 +274,7 @@ export function registerHuntCommand(program: Command): void {
           targetName: opts.name,
           objective: opts.objective,
           apiKeyEnv: opts.targetKeyEnv,
-          commanderModel: opts.model,
+          commanderModel: opts.commanderModel,
           operatorModel: opts.operatorModel,
           scoutModel: opts.scoutModel,
           maxOperators: opts.maxOperators,
@@ -453,7 +453,7 @@ export function registerHuntCommand(program: Command): void {
       const huntOptions: HuntOptions = {
         target,
         objective,
-        commanderModel: opts.model,
+        commanderModel: opts.commanderModel,
         operatorModel: opts.operatorModel,
         scoutModel: opts.scoutModel,
         maxOperators: intOr(opts.maxOperators, 6),

runners/extension/popup.js

@@ -1155,12 +1155,12 @@ function judgedCount(summary) {
  * Label for the report's "Evaluation Suite" field. A suite name is only
  * accurate when every one of its evaluators was actually selected — picking
  * a subset (or the catch-all "Custom Evaluators" bucket) is a hand-picked
- * list, not that suite, so it's labeled "Custom Suite" instead of borrowing
+ * list, not that suite, so it's labeled "Custom" instead of borrowing
  * a name that would overstate what actually ran.
  */
 function resolveSuiteLabel() {
   const suite = state.catalog?.suites?.find((s) => s.id === state.suiteId);
-  if (!suite) return "Custom Suite";
+  if (!suite) return "Custom";
   // Based on what actually completed, not what was selected — a run stopped
   // partway through a full-suite selection is not that suite either.
   const completedIds = new Set(
@@ -1169,7 +1169,7 @@ function resolveSuiteLabel() {
   const fullSuite =
     suite.evaluatorIds.length === completedIds.size &&
     suite.evaluatorIds.every((id) => completedIds.has(id));
-  return fullSuite ? suite.id : "Custom Suite";
+  return fullSuite ? suite.id : "Custom";
 }
 
 function buildReport() {

runners/sdk/src/types.ts

@@ -198,7 +198,7 @@ export interface RunResults {
   id: string;
   timestamp: string;
   targetName: string;
-  /** The evaluation suite that was run, or "Custom Suite" for a hand-picked selection. */
+  /** The evaluation suite that was run, or "Custom" for a hand-picked selection. */
   suiteId: string;
   targetKind: "agent" | "mcp";
   effort: Effort;