@withgauge/cli 0.11.0 → 0.11.2

package/README.md

@@ -98,6 +98,10 @@ immediate launches. They use the stored configuration and ask before spending
 (`--yes` skips the prompt for scripts).
 `gauge preference run <id> --scenario <scenarioId>` (repeatable) runs a subset;
 omit it and every scenario runs. Neither launch touches the cadence clock.
+Superusers can use `--org <slug>` to launch in any org. Their runs use PLATFORM
+funding by default, like Run Now in the web UI; `--bill-to-org` uses that org's
+credits instead. Other users always use org credits and can launch only in orgs
+they belong to.
 
 ## Optimizations: Gauge measures, you author
 
@@ -125,7 +129,7 @@ gauge optimizations trials add cl_1 \
 gauge optimizations watch cl_1
 gauge optimizations trial cl_1 R1                   # score, verdict, sessions
 gauge optimizations change cl_1 R1 --body           # the winning text, to paste
-gauge optimizations adopt cl_1 R1 --yes             # the new moving baseline
+gauge optimizations adopt cl_1 R1 --yes             # mark the winner; baseline stays fixed
 gauge optimizations complete cl_1 --yes
 ```
 

package/dist/index.js

@@ -6,7 +6,7 @@ var __export = (target, all) => {
 };
 
 // src/index.ts
-import { Command, CommanderError, Option as Option2 } from "commander";
+import { Command, CommanderError, Option as Option4 } from "commander";
 
 // src/commands/actions.ts
 var actions_exports2 = {};
@@ -1576,6 +1576,11 @@ var cancelRunResponseSchema = z17.object({
   cancelRequested: z17.boolean(),
   note: z17.string().optional()
 });
+var regradeRunResponseSchema = z17.object({
+  runId: z17.string(),
+  evalSetId: z17.string(),
+  criteria: z17.number().int()
+});
 var forkRunBodySchema = z17.object({
   prompt: z17.string().trim().min(1).max(2e4),
   turn: z17.number().int().min(0).optional(),
@@ -1718,6 +1723,7 @@ function totalRunsPerCycle(owners, samplesPerCycle) {
 
 // ../packages/api-schemas/src/ownedConfigurations.ts
 var sampleCountSchema = z19.number().int().min(1).max(50);
+var nextRunAtInputSchema = z19.string().regex(/^\d{4}-\d{2}-\d{2}/, "Next run must be a calendar date").nullable();
 var ownedCadenceSchema = z19.enum([
   "NONE",
   "DAILY",
@@ -1733,6 +1739,7 @@ var ownedConfigurationInputShape = {
   mcpRefs: assetRefListSchema.optional().default([]),
   connectionIds: z19.array(z19.string().min(1)).max(32).optional().default([]),
   cadence: ownedCadenceSchema.optional().default("NONE"),
+  nextRunAt: nextRunAtInputSchema.optional(),
   sampleCount: sampleCountSchema.optional().default(1)
 };
 var ownedConfigurationInputSchema = z19.object(ownedConfigurationInputShape).strict();
@@ -1745,6 +1752,7 @@ var ownedConfigurationPatchShape = {
   mcpRefs: assetRefListSchema.optional(),
   connectionIds: z19.array(z19.string().min(1)).max(32).optional(),
   cadence: ownedCadenceSchema.optional(),
+  nextRunAt: nextRunAtInputSchema.optional(),
   sampleCount: sampleCountSchema.optional()
 };
 var ownedConfigurationSchema = z19.object({
@@ -1757,7 +1765,8 @@ var ownedConfigurationSchema = z19.object({
   connectionIds: z19.array(z19.string()),
   cadence: ownedCadenceSchema,
   sampleCount: sampleCountSchema,
-  lastScheduledAt: z19.string().nullable()
+  lastScheduledAt: z19.string().nullable(),
+  nextRunAt: z19.string().nullable()
 });
 var visibilityScenarioSchema = z19.object({
   id: z19.string(),
@@ -1776,6 +1785,7 @@ var visibilityPromptSettingsInputShape = {
   mcpRefs: ownedConfigurationInputShape.mcpRefs,
   connectionIds: ownedConfigurationInputShape.connectionIds,
   cadence: ownedConfigurationInputShape.cadence,
+  nextRunAt: ownedConfigurationInputShape.nextRunAt,
   sampleCount: ownedConfigurationInputShape.sampleCount
 };
 var visibilityPromptSettingsPatchShape = {
@@ -1784,6 +1794,7 @@ var visibilityPromptSettingsPatchShape = {
   mcpRefs: ownedConfigurationPatchShape.mcpRefs,
   connectionIds: ownedConfigurationPatchShape.connectionIds,
   cadence: ownedConfigurationPatchShape.cadence,
+  nextRunAt: ownedConfigurationPatchShape.nextRunAt,
   sampleCount: ownedConfigurationPatchShape.sampleCount
 };
 var visibilityPromptSettingsSchema = z19.object({
@@ -1793,11 +1804,13 @@ var visibilityPromptSettingsSchema = z19.object({
   connectionIds: z19.array(z19.string()),
   cadence: ownedCadenceSchema,
   sampleCount: sampleCountSchema,
-  lastScheduledAt: z19.string().nullable()
+  lastScheduledAt: z19.string().nullable(),
+  nextRunAt: z19.string().nullable()
 });
 var runAgentsOverrideSchema = z19.array(agentConfigInputSchema).min(1, "Select at least one agent");
-var runEvalBodySchema = z19.object({}).strict();
+var runEvalBodySchema = z19.object({ billToOrg: z19.boolean().optional() }).strict();
 var runVisibilityBodySchema = z19.object({
+  billToOrg: z19.boolean().optional(),
   visibilityScenarioIds: z19.array(z19.string().min(1)).min(1).optional()
 }).strict();
 
@@ -2690,8 +2703,6 @@ var climbSettingsSchema = z29.object({
   maxRounds: z29.number().int().positive().default(5),
   sampleCount: z29.number().int().positive().default(1),
   creditCap: z29.number().int().nonnegative(),
-  exploreModel: z29.string().min(1).optional(),
-  autoAdopt: z29.boolean().default(false),
   targetScore: z29.number().min(0).max(1).default(1),
   // Narrow the subject's whole roster to one slot, for the baseline and every
   // trial alike. A subject rostering ten models otherwise multiplies every
@@ -2699,11 +2710,11 @@ var climbSettingsSchema = z29.object({
   // is not a comparison. `pinnedModel` empty means the agent's default model.
   pinnedAgent: schedulableAgentSchema.optional(),
   pinnedModel: z29.string().optional(),
-  maxTrialsPerRound: z29.number().int().min(1).max(3).default(3),
+  maxTrialsPerRound: z29.number().int().min(1).max(10).default(3),
   // The round a continuation started at; the no-improvement counter
   // restarts here.
   continuedAtRound: z29.number().int().nonnegative().optional()
-}).strict().refine((value) => value.pinnedAgent !== void 0 || !value.pinnedModel, {
+}).strip().refine((value) => value.pinnedAgent !== void 0 || !value.pinnedModel, {
   path: ["pinnedAgent"],
   message: "Pin an agent with the model"
 });
@@ -2727,7 +2738,6 @@ var CLIMB_ACTIVITY_KINDS = [
   "EXCLUDED",
   "LAUNCHED",
   "SETTLED",
-  "CONFIRMING",
   "DONE",
   "FAILED",
   "DIRECTION",
@@ -2882,9 +2892,9 @@ var optimizationSummarySchema = z30.object({
   maxRounds: z30.number().int(),
   /** Starting baseline score, 0..1. */
   before: z30.number().nullable(),
-  /** Score of the adopted tip, 0..1. */
+  /** Score of the marked winner, or best BETTER trial, 0..1. */
   after: z30.number().nullable(),
-  /** Percentage-point change from the starting baseline to the adopted tip. */
+  /** Percentage-point change from the baseline to the winner or best BETTER trial. */
   delta: z30.number().nullable(),
   /** Score after each round, baseline first. */
   path: z30.array(z30.number()),
@@ -2907,7 +2917,7 @@ var planShape = {
   effort: optimizationEffortSchema.optional(),
   sampleCount: z30.number().int().min(1).max(100).optional(),
   maxRounds: z30.number().int().min(1).max(100).optional(),
-  maxTrialsPerRound: z30.number().int().min(1).max(3).optional()
+  maxTrialsPerRound: z30.number().int().min(1).max(10).optional()
 };
 function refinePlan(value, ctx) {
   if (value.effort && (value.sampleCount || value.maxRounds))
@@ -2929,7 +2939,6 @@ var startOptimizationBodySchema = z30.object({
   name: z30.string().trim().min(1).max(200).optional(),
   ...planShape,
   creditCap: z30.number().int().nonnegative().optional(),
-  autoAdopt: z30.boolean().default(true),
   targetScore: z30.number().min(0).max(1).optional(),
   target: targetSchema.extend({ agent: schedulableAgentSchema })
 }).strict().superRefine(refinePlan);
@@ -3010,7 +3019,9 @@ var trialSchema = z30.object({
   climbId: z30.string(),
   roundNumber: z30.number().int(),
   key: z30.string(),
+  /** @deprecated All trials branch from the baseline. */
   parentTrialId: z30.string().nullable(),
+  /** @deprecated Always the baseline key for non-baseline trials. */
   parentKey: z30.string().nullable(),
   title: z30.string(),
   hypothesis: z30.string().nullable(),
@@ -3022,8 +3033,8 @@ var trialSchema = z30.object({
   baselineScore: z30.number().nullable(),
   /** Judged sessions behind `score`. */
   sessions: z30.number().int(),
-  explore: z30.boolean(),
   isBaseline: z30.boolean(),
+  /** The user marked this trial as the winner; it does not move the baseline. */
   adopted: z30.boolean(),
   adoptedAt: z30.string().nullable(),
   outcome: z30.string().nullable(),
@@ -3059,9 +3070,7 @@ var optimizationDetailSchema = optimizationSummarySchema.extend({
     maxRounds: z30.number().int(),
     maxTrialsPerRound: z30.number().int(),
     creditCap: z30.number().int(),
-    autoAdopt: z30.boolean(),
     targetScore: z30.number(),
-    exploreModel: z30.string().nullable(),
     pinnedAgent: z30.string().nullable(),
     pinnedModel: z30.string().nullable(),
     continuedAtRound: z30.number().int().nullable()
@@ -3138,9 +3147,8 @@ var resolveChatResponseSchema = z30.object({
   href: z30.string()
 });
 var continueOptimizationBodySchema = z30.object({
-  fromTrialId: z30.string().min(1),
   rounds: z30.number().int().min(1).max(20).default(1),
-  maxTrialsPerRound: z30.number().int().min(1).max(3).optional()
+  maxTrialsPerRound: z30.number().int().min(1).max(10).optional()
 }).strict();
 var patchOptimizationBodySchema = z30.object({
   ...planPatchShape,
@@ -5182,6 +5190,7 @@ __export(evals_exports2, {
   register: () => register11
 });
 import { readFileSync as readFileSync6 } from "fs";
+import { Option as Option2 } from "commander";
 
 // src/lib/confirm.ts
 import { createInterface as createInterface2 } from "readline";
@@ -5203,6 +5212,11 @@ async function confirmOrAbort(question, yes) {
   }
 }
 
+// src/lib/staff.ts
+function staffCommandsVisible(config = readConfig()) {
+  return config.staff === true;
+}
+
 // src/lib/ownedConfig.ts
 function parseCadence(raw) {
   const cadence = (raw ?? "NONE").trim().toUpperCase();
@@ -5590,23 +5604,30 @@ function register11(program2) {
     if (resolveOutput(cmd) !== "json") console.log(`Deleted eval set ${id}`);
   });
   group.command("run <id>").description(
-    "Run an eval set now with its stored configuration (ORG funding, credit-gated)"
-  ).option("--yes", "skip the confirmation").action(async (id, opts, cmd) => {
-    const org = resolveOrg(cmd);
-    await confirmOrAbort(
-      "Run this eval's stored configuration now?",
-      opts.yes
-    );
-    const result = await createClient().post(
-      `${evalSetsPath(org)}/${encodeURIComponent(id)}/run`,
-      {}
-    );
-    if (resolveOutput(cmd) === "json") printJson(result);
-    else
-      console.log(
-        `Started ${result.runIds.length} run${result.runIds.length === 1 ? "" : "s"} in batch ${result.batchIds[0] ?? ""}`
+    staffCommandsVisible() ? "Run an eval set now with its stored configuration (superusers run without billing the org by default)" : "Run an eval set now with its stored configuration"
+  ).addOption(
+    new Option2(
+      "--bill-to-org",
+      "bill this run to the org's credits instead (superusers)"
+    ).hideHelp(!staffCommandsVisible())
+  ).option("--yes", "skip the confirmation").action(
+    async (id, opts, cmd) => {
+      const org = resolveOrg(cmd);
+      await confirmOrAbort(
+        "Run this eval's stored configuration now?",
+        opts.yes
       );
-  });
+      const result = await createClient().post(
+        `${evalSetsPath(org)}/${encodeURIComponent(id)}/run`,
+        opts.billToOrg ? { billToOrg: true } : {}
+      );
+      if (resolveOutput(cmd) === "json") printJson(result);
+      else
+        console.log(
+          `Started ${result.runIds.length} run${result.runIds.length === 1 ? "" : "s"} in batch ${result.batchIds[0] ?? ""}`
+        );
+    }
+  );
   group.command("runs <id>").description("The eval set's run log with per-criterion verdicts").option("--limit <n>", "max runs (default 100)").action(async (id, opts, cmd) => {
     const org = resolveOrg(cmd);
     const res = await createClient().get(
@@ -5723,7 +5744,7 @@ function printDetail3(e) {
 }
 function register12(program2) {
   const group = program2.command("experiments").description(
-    "Author, launch, and read Fixer content experiments (legacy: orgs with Optimizations answer 409 on writes; use `gauge optimizations`)"
+    "Author, launch, and read content experiments (legacy: orgs with Optimizations reject writes; use `gauge optimizations`)"
   );
   group.command("list").description("List the org's experiments").option(
     "--status <statuses>",
@@ -5769,13 +5790,13 @@ function register12(program2) {
       );
   });
   group.command("create").description(
-    "Open (or resume) an experiment draft and start the Fixer plan round"
+    "Open (or resume) an experiment draft and start the plan round"
   ).requiredOption("--run <runId>", "source run id").requiredOption(
     "--boundary <seq>",
     "opaque boundarySeq from `experiments boundaries`"
   ).option("--problem <text>", "problem statement (derived if omitted)").option(
     "--mode <mode>",
-    "who writes the rewrites: propose (default, Fixer proposes directions then writes each), direct (one --change instruction, one write), manual (no generation \u2014 the rewrite is seeded with the page the agent read)",
+    "who writes the rewrites: propose (default, directions are proposed then each is written), direct (one --change instruction, one write), manual (no generation \u2014 the rewrite is seeded with the page the agent read)",
     "propose"
   ).option(
     "--change <text>",
@@ -5959,13 +5980,13 @@ function register12(program2) {
       }
     }
   );
-  group.command("replan <id>").description("Discard the directions and have the Fixer propose fresh ones").action(async (id, _opts, cmd) => {
+  group.command("replan <id>").description("Discard the directions and propose fresh ones").action(async (id, _opts, cmd) => {
     const org = resolveOrg(cmd);
     await createClient().post(`${itemPath(org, id)}/replan`, {});
     if (resolveOutput(cmd) !== "json")
       console.log("Replanning \u2014 follow with `gauge experiments watch`");
   });
-  group.command("write <id>").description("Have the Fixer write rewrites for directions that lack one").action(async (id, _opts, cmd) => {
+  group.command("write <id>").description("Write rewrites for directions that lack one").action(async (id, _opts, cmd) => {
     const org = resolveOrg(cmd);
     const result = await createClient().post(
       `${itemPath(org, id)}/write`,
@@ -6411,7 +6432,7 @@ function register15(program2) {
   });
   group.command("add").description(
     "Add a server, or append a version to an existing --name (versions are immutable)"
-  ).requiredOption("--name <handle>", "our handle (the agent config key)").option("--command <cmd>", "stdio launch command, e.g. npx").option(
+  ).requiredOption("--name <handle>", "the handle (the agent config key)").option("--command <cmd>", "stdio launch command, e.g. npx").option(
     "--arg <arg>",
     "stdio argument (repeatable)",
     collect,
@@ -6484,7 +6505,7 @@ function register15(program2) {
     }
   );
   group.command("rm <ref>").description(
-    "Soft-delete a server (handle stays reserved; finished runs keep their snapshot)"
+    "Remove a server; the handle stays reserved and finished runs keep their snapshot"
   ).action(async (ref, _opts, cmd) => {
     const org = resolveOrg(cmd);
     const result = await createClient().delete(serverPath(org, ref));
@@ -6883,7 +6904,6 @@ function trialBodyFromOpts(opts, read = readText2, walk) {
   }
   if (!opts.title) throw new CliError("usage", "--title is required");
   const lever = opts.lever?.toUpperCase();
-  const kind = opts.kind?.toUpperCase();
   let intervention;
   let files;
   if (opts.page || opts.bodyFile) {
@@ -6893,12 +6913,6 @@ function trialBodyFromOpts(opts, read = readText2, walk) {
       kind: "REWRITE_DOCS",
       after: { pages: [{ url: opts.page, body: read(opts.bodyFile) }] }
     };
-  } else if (opts.removeSkill) {
-    intervention = {
-      kind: "REMOVE_SKILL",
-      before: { skillVersionId: opts.removeSkill },
-      after: {}
-    };
   } else if (opts.editSkill) {
     if (!opts.skillDir)
       throw new CliError(
@@ -6911,21 +6925,10 @@ function trialBodyFromOpts(opts, read = readText2, walk) {
       after: { skillVersionId: opts.editSkill }
     };
     files = readSkillDir(opts.skillDir, read, walk);
-  } else if (opts.skillVersion) {
-    if (opts.skillDir)
-      throw new CliError(
-        "usage",
-        "--skill-dir edits a skill: pass --edit-skill <id>, not --skill-version"
-      );
-    intervention = {
-      kind: kind ?? "ADD_SKILL",
-      after: { skillVersionId: opts.skillVersion },
-      ...kind === "EDIT_SKILL" ? { before: { skillVersionId: opts.skillVersion } } : {}
-    };
   } else {
     throw new CliError(
       "usage",
-      "describe the change: --skill-version <id>, --edit-skill <id> --skill-dir <path>, --remove-skill <id>, --page <url> --body-file <path>, or --file <trial.json>"
+      "describe the change: --edit-skill <id> --skill-dir <path>, --page <url> --body-file <path>, or --file <trial.json>"
     );
   }
   const trial = {
@@ -6944,18 +6947,6 @@ function trialBodyFromOpts(opts, read = readText2, walk) {
   };
 }
 var TRIAL_TEMPLATES = {
-  ADD_SKILL: {
-    trial: {
-      title: "Add the troubleshooting skill",
-      hypothesis: "The agent misses the retry rule without it",
-      lever: "SKILL",
-      intervention: {
-        kind: "ADD_SKILL",
-        after: { skillVersionId: "sv_REPLACE_ME" }
-      },
-      authorMode: "MANUAL"
-    }
-  },
   EDIT_SKILL: {
     trial: {
       title: "Name the env var in the first line",
@@ -6972,19 +6963,6 @@ var TRIAL_TEMPLATES = {
       { path: "SKILL.md", body: "# Skill\n\nThe complete rewritten file.\n" }
     ]
   },
-  REMOVE_SKILL: {
-    trial: {
-      title: "Drop the legacy skill",
-      hypothesis: "It contradicts the quickstart",
-      lever: "SKILL",
-      intervention: {
-        kind: "REMOVE_SKILL",
-        before: { skillVersionId: "sv_REPLACE_ME" },
-        after: {}
-      },
-      authorMode: "MANUAL"
-    }
-  },
   REWRITE_DOCS: {
     trial: {
       title: "Move the auth note above the example",
@@ -7041,12 +7019,11 @@ function trialRow(t) {
     title: t.title.length > 48 ? `${t.title.slice(0, 45)}...` : t.title,
     surface: t.lever ?? "",
     change: t.kind,
-    "builds on": t.parentKey ?? "",
     status: t.status,
     score: pct(t.score),
     result: t.verdict ?? "",
     sessions: t.sessions,
-    adopted: t.adopted ? "yes" : "",
+    winner: t.adopted ? "yes" : "",
     id: t.id
   };
 }
@@ -7066,7 +7043,7 @@ function printDetail4(o) {
     `target:    ${o.settings.pinnedAgent ?? "(subject roster)"}${o.settings.pinnedModel ? `:${o.settings.pinnedModel}` : ""}`
   );
   console.log(
-    `plan:      ${o.settings.sampleCount} session${o.settings.sampleCount === 1 ? "" : "s"} per trial, up to ${o.settings.maxRounds} rounds, auto-adopt ${o.settings.autoAdopt ? "on" : "off"}`
+    `plan:      ${o.settings.sampleCount} session${o.settings.sampleCount === 1 ? "" : "s"} per trial, up to ${o.settings.maxRounds} rounds`
   );
   console.log(
     `credits:   ${o.creditsSpent} spent; ${o.settings.creditCap} estimated (not a spending cap)`
@@ -7089,9 +7066,8 @@ function printTrial(t) {
   console.log(`title:      ${t.title}`);
   if (t.hypothesis) console.log(`hypothesis: ${t.hypothesis}`);
   console.log(`change:     ${t.kind}${t.lever ? ` on ${t.lever}` : ""}`);
-  console.log(`builds on:  ${t.parentKey ?? "\u2014"}`);
   console.log(
-    `status:     ${t.status}${t.verdict ? ` \xB7 ${t.verdict}` : ""}${t.adopted ? " \xB7 adopted" : ""}`
+    `status:     ${t.status}${t.verdict ? ` \xB7 ${t.verdict}` : ""}${t.adopted ? " \xB7 winner" : ""}`
   );
   console.log(
     `score:      ${pct(t.baselineScore)} \u2192 ${pct(t.score)} over ${t.sessions} session${t.sessions === 1 ? "" : "s"}`
@@ -7165,7 +7141,7 @@ function addPlanOptions(command) {
 }
 function register19(program2) {
   const group = program2.command("optimizations").alias("optimize").description(
-    "Start, steer, and read optimizations: Gauge changes one owned surface per trial and adopts what improves an eval or preference prompt"
+    "Start, steer, and read optimizations: Gauge changes one owned surface per trial and reports the best measured change against the baseline"
   );
   group.command("list").description("List the org's optimizations, newest activity first").option(
     "--status <statuses>",
@@ -7226,10 +7202,10 @@ function register19(program2) {
         "Measure a baseline; author and explicitly launch candidate rounds afterward"
       )
     )
-  ).option("--name <name>", "climb name (default: from the subject)").option(
+  ).option("--name <name>", "optimization name (default: from the subject)").option(
     "--credit-cap <n>",
     "record this credit estimate instead (not an enforced cap)"
-  ).option("--target-score <0-1>", "stop once the score reaches this").option("--no-auto-adopt", "keep improvements pending until you adopt them").option("--yes", "skip the confirmation").action(
+  ).option("--target-score <0-1>", "stop once the score reaches this").option("--yes", "skip the confirmation").action(
     async (opts, cmd) => {
       const org = resolveOrg(cmd);
       if (!opts.target)
@@ -7258,7 +7234,6 @@ function register19(program2) {
         ...plan,
         name: opts.name,
         creditCap,
-        autoAdopt: opts.autoAdopt,
         targetScore
       });
       if (resolveOutput(cmd) === "json") printJson(detail);
@@ -7392,9 +7367,8 @@ function register19(program2) {
         result.alreadyLaunched ? "Round already launched." : `Launched ${result.launched.length} trials in round ${result.roundNumber}.`
       );
   });
-  group.command("continue <id>").description("Continue a finished optimization from a settled trial").requiredOption(
-    "--from <trialId>",
-    "starting trial id or key, e.g. B0 or R1.1"
+  group.command("continue <id>").description(
+    "Add rounds to a finished optimization against its fixed baseline"
   ).option("--rounds <n>", "additional rounds (1\u201320, default 1)").option("--trials-per-round <n>", "maximum candidates per round (1\u20133)").option("--yes", "skip the confirmation").action(
     async (id, opts, cmd) => {
       const rounds = parseCount(opts.rounds, "--rounds", 20) ?? 1;
@@ -7404,12 +7378,12 @@ function register19(program2) {
         3
       );
       await confirmOrAbort(
-        `Continue ${id} from ${opts.from} for up to ${rounds} additional rounds? Saved trials still require launch.`,
+        `Continue ${id} for up to ${rounds} additional rounds? Saved trials still require launch.`,
         opts.yes
       );
       const detail = await createClient().post(
         `${itemPath2(resolveOrg(cmd), id)}/continue`,
-        { fromTrialId: opts.from, rounds, maxTrialsPerRound }
+        { rounds, maxTrialsPerRound }
       );
       if (resolveOutput(cmd) === "json") printJson(detail);
       else printDetail4(detail);
@@ -7461,7 +7435,7 @@ function register19(program2) {
     if (resolveOutput(cmd) === "json") printJson({ deleted: id });
     else console.log(`Deleted ${id}.`);
   });
-  group.command("complete <id>").description("Complete an optimization: it reads as Finished, not canceled").option("--adopt <trialId>", "adopt this settled trial as the winner first").option("--yes", "skip the confirmation").action(
+  group.command("complete <id>").description("Complete an optimization: it reads as Finished, not canceled").option("--adopt <trialId>", "mark this settled trial as the winner").option("--yes", "skip the confirmation").action(
     async (id, opts, cmd) => {
       const org = resolveOrg(cmd);
       await confirmOrAbort(
@@ -7525,15 +7499,12 @@ function register19(program2) {
     "--file <path>",
     "authored trial JSON, as `trials template` prints it: one {trial, files?} or an array; '-' for stdin"
   ).option("--pick <n>", "which trial to take from an array (default 1)").option("--title <text>", "what the change is").option("--hypothesis <text>", "why it should help").option("--lever <lever>", "DOCS, MARKETING_SITE, SKILL, SDK, CLI").option(
-    "--kind <kind>",
-    "override the change kind the other flags imply; rarely needed"
-  ).option("--skill-version <id>", "skill version to add").option(
     "--edit-skill <id>",
     "skill version to rewrite (with --skill-dir); see `gauge optimizations skill-files <id>`"
   ).option(
     "--skill-dir <path>",
     "local directory holding the complete edited bundle, SKILL.md included"
-  ).option("--remove-skill <id>", "skill version to remove from the baseline").option("--page <url>", "docs page to rewrite (with --body-file)").option("--body-file <path>", "the rewritten page body; '-' for stdin").option("--fork-run <runId>", "fork-at-fetch source run for a docs rewrite").option(
+  ).option("--page <url>", "docs page to rewrite (with --body-file)").option("--body-file <path>", "the rewritten page body; '-' for stdin").option("--fork-run <runId>", "fork-at-fetch source run for a docs rewrite").option(
     "--fork-boundary <seq>",
     "fork-at-fetch boundary from `gauge experiments boundaries`"
   ).option(
@@ -7563,10 +7534,7 @@ function register19(program2) {
         );
     }
   );
-  trials.command("template").description("Print a trial JSON skeleton for `trials add --file`").requiredOption(
-    "--kind <kind>",
-    "ADD_SKILL, EDIT_SKILL, REMOVE_SKILL, or REWRITE_DOCS"
-  ).action((opts) => {
+  trials.command("template").description("Print a trial JSON skeleton for `trials add --file`").requiredOption("--kind <kind>", "EDIT_SKILL or REWRITE_DOCS").action((opts) => {
     printJson(trialTemplate(opts.kind));
   });
   trials.command("run <id> <trialId>").description("Run a draft trial now (spends credits)").option("--yes", "skip the confirmation").action(
@@ -7615,12 +7583,12 @@ function register19(program2) {
     }
   );
   group.command("adopt <id> <trialId>").description(
-    "Make a settled trial the moving baseline (the eval itself is unchanged)"
+    "Mark a settled trial as the winner without changing the baseline"
   ).option("--yes", "skip the confirmation").action(
     async (id, trialId, opts, cmd) => {
       const org = resolveOrg(cmd);
       await confirmOrAbort(
-        `Adopt ${trialId} on ${id}? Later trials build on it.`,
+        `Mark ${trialId} as the winner on ${id}?`,
         opts.yes
       );
       const detail = await createClient().post(
@@ -7628,14 +7596,11 @@ function register19(program2) {
         {}
       );
       if (resolveOutput(cmd) === "json") printJson(detail);
-      else
-        console.log(
-          `Adopted ${trialId}; baseline is now ${pct(detail.after)}`
-        );
+      else console.log(`Marked ${trialId} as the winner.`);
     }
   );
   group.command("skill-files <versionId>").description(
-    "The text files of a skill version, the base an EDIT_SKILL trial edits"
+    "The text files of a skill version, the base a skill-edit trial edits"
   ).action(async (versionId, _opts, cmd) => {
     const org = resolveOrg(cmd);
     const res = await createClient().get(
@@ -7649,7 +7614,7 @@ function register19(program2) {
       }
   });
   group.command("page <id>").description(
-    "The baseline body of a docs page, the base a REWRITE_DOCS trial edits"
+    "The baseline body of a docs page, the base a docs-rewrite trial edits"
   ).requiredOption("--url <url>", "the page URL").action(async (id, opts, cmd) => {
     const org = resolveOrg(cmd);
     const res = await createClient().get(
@@ -8510,6 +8475,21 @@ function register25(program2) {
     }
     printInsight(insight);
   });
+  runs.command("regrade <id>").description(
+    "Re-judge the run against its eval set's current criteria (no new session)"
+  ).action(async (id, _opts, command) => {
+    const result = await createClient().post(
+      `${runPath(id)}/regrade`
+    );
+    if (resolveOutput(command) === "json") {
+      printJson(result);
+      return;
+    }
+    console.log(
+      `Re-judging run ${result.runId} against ${result.criteria} ${result.criteria === 1 ? "criterion" : "criteria"} of eval ${result.evalSetId}.`
+    );
+    console.log("Verdicts land on the run; `gauge runs get <id>` to check.");
+  });
   runs.command("watch <id>").description(
     "Stream normalized trace events live (SSE; exits when the run finishes)"
   ).action(async (id, _opts, command) => {
@@ -8682,7 +8662,7 @@ function register26(program2) {
         `  ${skill2.name}@${v.label}  ${v.contentDigest.slice(0, 19)}  ${v.createdAt}`
       );
   });
-  group.command("add <reference>").description("Ingest a skill from owner/repo[/skill][#ref] or a GitHub URL").option("--as <handle>", "our handle (default: the directory basename)").option(
+  group.command("add <reference>").description("Ingest a skill from owner/repo[/skill][#ref] or a GitHub URL").option("--as <handle>", "the handle (default: the directory basename)").option(
     "--label <label>",
     "version label (a typed label always mints; omitted dedups by digest)"
   ).option("--skill <name>", "select one skill from a source holding several").option("--note <text>", "note shown on the index").option(
@@ -8747,7 +8727,7 @@ function register26(program2) {
       );
   });
   group.command("rm <ref>").description(
-    "Soft-delete a skill (handle stays reserved; finished runs keep their snapshot)"
+    "Remove a skill; the handle stays reserved and finished runs keep their snapshot"
   ).action(async (ref, _opts, cmd) => {
     const org = resolveOrg(cmd);
     const result = await createClient().delete(skillPath(org, ref));
@@ -9071,6 +9051,7 @@ __export(visibility_exports2, {
   register: () => register31,
   visibilityPath: () => visibilityPath
 });
+import { Option as Option3 } from "commander";
 function visibilityPath(org) {
   return `/api/v1/orgs/${encodeURIComponent(org)}/visibility-prompts`;
 }
@@ -9298,6 +9279,11 @@ function registerPreferenceGroup(group) {
     "run only this scenario (repeatable; default: all)",
     collect,
     []
+  ).addOption(
+    new Option3(
+      "--bill-to-org",
+      "bill this run to the org's credits instead (superusers)"
+    ).hideHelp(!staffCommandsVisible())
   ).option("--yes", "skip the confirmation").action(
     async (id, opts, cmd) => {
       const org = resolveOrg(cmd);
@@ -9308,6 +9294,7 @@ function registerPreferenceGroup(group) {
       const result = await createClient().post(
         `${visibilityPath(org)}/${encodeURIComponent(id)}/run`,
         {
+          ...opts.billToOrg ? { billToOrg: true } : {},
           visibilityScenarioIds: opts.scenario.length > 0 ? opts.scenario : void 0
         }
       );
@@ -9424,11 +9411,6 @@ function register31(program2) {
   registerPreferenceGroup(legacy);
 }
 
-// src/lib/staff.ts
-function staffCommandsVisible(config = readConfig()) {
-  return config.staff === true;
-}
-
 // src/index.ts
 var program = new Command();
 program.name("gauge").description(
@@ -9437,7 +9419,7 @@ program.name("gauge").description(
   "--org <org>",
   "organization slug or public id (default: `gauge config get defaultOrg`)"
 ).addOption(
-  new Option2("-o, --output <format>", "output format").choices([...OUTPUT_FORMATS]).default("table")
+  new Option4("-o, --output <format>", "output format").choices([...OUTPUT_FORMATS]).default("table")
 );
 for (const group of [
   instructions_exports,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@withgauge/cli",
-  "version": "0.11.0",
+  "version": "0.11.2",
   "description": "Gauge customer CLI — Agent Preference, evals, runs, and organization settings over the Gauge API",
   "private": false,
   "author": "Gauge <support@withgauge.com> (https://withgauge.com)",
@@ -22,7 +22,9 @@
   "scripts": {
     "build": "tsup",
     "dev": "tsx src/index.ts",
-    "test": "tsx --test src/**/*.test.ts",
+    "prepublishOnly": "node scripts/validate-release.mjs",
+    "release:validate": "node scripts/validate-release.mjs",
+    "test": "tsx --test src/**/*.test.ts scripts/*.test.mjs",
     "typecheck": "tsc --noEmit"
   },
   "dependencies": {