@evolvingmachines/evolve 0.0.67 → 0.0.69

package/dist/{types-84M4J_Vm.d.ts → types-DGx_qo9M.d.ts}

@@ -450,8 +450,8 @@ interface Rubric {
  * openrouter/deepseek/deepseek-v4.1-flash at its per-model effort (high)
  * over the platform's default analyze rubric and its default prompt body.
  *
- * The analyzer always runs the claude-code harness (Harbor's default analyze
- * agent) in its own sealed sandbox — on the provider `sandbox_provider`
+ * The analyzer runs on the `agent` harness (default claude, Harbor's claude-code)
+ * in its own sealed sandbox — on the provider `sandbox_provider`
  * names, or the platform's analysis default when it names none; its spend is
  * capped per analysis and metered as its own line, never blended into the
  * trial's own bill.
@@ -460,31 +460,36 @@ interface Rubric {
  * exact names — `n_concurrent` (`-n/--n-concurrent`), `passing` / `failing`,
  * `n_trials` (`-l/--n-trials`; their cli/analyze.py:278-290). All omitted
  * is every analyzable trial, as wide as the organization's
- * `max_concurrent_analyses` allows. Harbor's `-a/--agent`, `--job-name`,
+ * `max_concurrent_analyses` allows. Harbor's `--job-name`,
  * `-o/--jobs-dir`, `-k/--n-attempts` and the local-runner kwargs are not
  * on this surface; the contract (`AnalyzeConfigInput` in spec/openapi.yaml)
  * records each with its reason.
  */
 interface AnalyzeConfigInput {
-    /**
-     * Model the analyzer agent runs — Harbor's `--model`. The default is
-     * openrouter/deepseek/deepseek-v4.1-flash on this platform's claude
-     * roster (DeepSeek V4.1 Flash served through OpenRouter, at its default
-     * effort high — the owner's ruling 2026-09-10: far more parallel capacity
+    /** Harbor's `-a/--agent`, spelled as the arms spell it (claude, not claude-code) so one set of names covers arms and reviewers; omitted: claude. The agents: `GET /api/meta` `analyze.agents`; a retired one (`retired_agents`) is refused `agent_retired`. */
+    agent?: string;
+    /**
+     * Model the analyzer agent runs — Harbor's `--model`. Omitted, the agent's
+     * default: the platform's pick, openrouter/deepseek/deepseek-v4.1-flash
+     * (`GET /api/meta` `analyze.default_model`), on every agent whose roster
+     * carries it, else that agent's own default (`analyze.agents[].default_model`,
+     * the model the SDK runs that agent on when none is named). The platform's
+     * pick is DeepSeek V4.1 Flash served through OpenRouter, at its default
+     * effort high (the owner's ruling 2026-09-10: far more parallel capacity
      * through OpenRouter's provider pool than one pinned Fireworks host) — a
      * recorded deviation from Harbor's default analyze model (their
      * cli/analyze.py `claude-haiku-4-5`): analysis is input-dominated, and
      * this is the roster's intelligence-per-input-dollar pick; `glm-5.3-flash`
      * (at max, the effort its published scores use) and `haiku` stay on the
-     * roster as alternatives, `glm-5.3` to escalate, and the same model on
-     * its Fireworks route, `fireworks/deepseek-v4.1-flash`, is a further
-     * option (the OpenRouter id stays the default). The value speaks
-     * the same vocabulary as `agents[].model_name`: either advertised
-     * spelling is accepted and stored AS GIVEN (the default is the roster
-     * alias), the wire id is resolved only when the analyzer runs, and every
-     * stored analysis serves the spelling it was created under. Must be on
-     * the claude roster (`GET /api/meta`, `agents[].models`); anything else
-     * is refused at accept (`invalid_input`, roster in the message).
+     * claude roster as alternatives, `glm-5.3` to escalate, and the same model
+     * on its Fireworks route, `fireworks/deepseek-v4.1-flash`, is a further
+     * option (the OpenRouter id stays the pick). The value speaks the same
+     * vocabulary as `agents[].model_name`: either advertised spelling is
+     * accepted and stored AS GIVEN (the default is the roster alias), the wire
+     * id is resolved only when the analyzer runs, and every stored analysis
+     * serves the spelling it was created under. A named model must be on the
+     * agent's roster (`GET /api/meta`, `analyze.agents[].models`); anything
+     * else is refused at accept (`invalid_input`, roster in the message).
      */
     model_name?: string;
     rubric?: Rubric;
@@ -511,15 +516,14 @@ interface AnalyzeConfigInput {
     prompt?: string;
     /**
      * Reasoning effort the analyzer runs at — the platform's `agents[].
-     * reasoning_effort` vocabulary applied to the analyzer, which IS the
-     * claude harness: the accepted values are `GET /api/meta`'s
-     * `analyze.reasoning_efforts`, an unknown value is refused
+     * reasoning_effort` vocabulary applied to the analyzer's agent: the accepted
+     * values are `GET /api/meta`'s `analyze.agents[].reasoning_efforts`, an unknown value is refused
      * `invalid_input` exactly as an arm's is. Omitted, the PER-MODEL default
-     * applies (`analyze.models[].default_reasoning_effort`: high on
+     * applies (`analyze.agents[].models[].default_reasoning_effort`: high on
      * openrouter/deepseek/deepseek-v4.1-flash, the default model — DeepSeek's
      * own documented default, the owner's ruling 2026-09-10; max on
      * glm-5.3-flash — the platform's ruling 2026-09-08, the effort its
-     * published scores use; the claude harness default elsewhere). The
+     * published scores use; the agent's own default elsewhere, none where it takes none). The
      * effort is always passed to the analyzer explicitly and
      * recorded on the analysis (`TrialAnalysis.reasoning_effort`). A hosted
      * extension: Harbor's analyze has no effort option; this is the run
@@ -596,10 +600,12 @@ interface AnalyzeConfigInput {
  * the day, resolved at accept and stored, so the record always states the
  * policy it executes (same law as RetryConfig). Echoed on the job body when
  * the job was created with `analyze`; each analysis additionally carries the
- * exact policy IT ran under (`Trial.analysis.model_name` / `.rubric` /
- * `.prompt`), which a later manual re-analysis may have changed.
+ * exact policy IT ran under (`Trial.analysis.agent` / `.model_name` /
+ * `.rubric` / `.prompt`), which a later manual re-analysis may have changed.
  */
 interface AnalyzeConfig {
+    /** The agent this policy's analyses run on — as named, or the default (claude). */
+    agent: string;
     model_name: string;
     rubric: Rubric;
     /** The caller's prompt template as stored; null = the platform's default analyze body. */
@@ -607,10 +613,10 @@ interface AnalyzeConfig {
     /**
      * The effort this policy's analyses run at. Named at create it is served
      * as stored; when the create named none, this echoes the per-model
-     * default of the day for `model_name` — the value the next enqueue under
-     * this policy stamps (the same nuance as `sandbox_provider` below).
+     * default of the day for `model_name` on `agent` — the value the next enqueue under
+     * this policy stamps (the same nuance as `sandbox_provider` below). Null when the agent takes no effort.
      */
-    reasoning_effort: string;
+    reasoning_effort: string | null;
     /**
      * The provider this policy's analyses run on. Named at create it is served
      * as stored, forever. When the create named none, this echoes the
@@ -1485,9 +1491,12 @@ interface AnalysisFailure {
      * `timeout` (the analyzer's run budget ran out with no valid analysis.json
      * — the file missing, or a partial one that failed validation, its reasons
      * in the message — never re-run: a timeout is deterministic; the message
-     * names the budget, the seconds used and the exit code), or an
-     * infrastructure stage of the analyzer run (`mint_key`, `boot`,
-     * `harness_install`, `agent`, `artifact_read`, `lease_expired`, ...).
+     * names the budget, the seconds used and the exit code), `agent_retired`
+     * (the job's stored analysis agent was retired after the job was created:
+     * recorded when the trial settles, nothing run or charged, never re-run; the
+     * message names the replacement), or an infrastructure stage of the analyzer
+     * run (`mint_key`, `boot`, `harness_install`, `agent`, `artifact_read`,
+     * `lease_expired`, ...).
      */
     phase: string;
     message: string;
@@ -1523,11 +1532,13 @@ interface TrialAnalysis {
      * `failed`, never left `running` forever.
      */
     status: AnalysisStatus;
+    /** The agent THIS analysis ran on; every analysis from before the choice existed ran on claude. */
+    agent: string;
     model_name: string;
     /**
      * The reasoning effort THIS analysis ran at — passed to the analyzer
-     * explicitly, so it is what the model was asked for. Null only on
-     * analyses recorded before the effort was stamped.
+     * explicitly, so it is what the model was asked for. Null when its agent
+     * takes no effort, or on analyses recorded before the effort was stamped.
      */
     reasoning_effort: string | null;
     rubric: Rubric;
@@ -2268,10 +2279,11 @@ interface CheckRow {
     name: string;
     status: CheckStatus;
     source: CheckSource;
+    agent: string;
     /** The checker's model. */
     model_name: string;
-    /** The effort every task's checker ran at. */
-    reasoning_effort: string;
+    /** The effort every task's checker ran at; null when its agent takes none. */
+    reasoning_effort: string | null;
     sandbox_provider: EvalSandboxProvider;
     /** The owning organization's slug (every check has one). */
     org: string;
@@ -4548,8 +4560,10 @@ interface AnalysesClient {
      * feed's species-blind events door, which answers `trial_not_found`.
      */
     list(options?: ListAnalysesOptions): AnalysisList;
-    /** The defaults an analysis runs under when its config names nothing (GET /api/analyses/defaults): model, effort, provider, rubric and the unrendered prompt template. */
-    defaults(): Promise<AnalyzeDefaults>;
+    /** The defaults an analysis runs under when its config names nothing (GET /api/analyses/defaults): agent, model, effort, provider, rubric and the unrendered prompt template. `{ agent }` reads what that agent runs under — its default model and the effort it takes; an unknown name is refused `invalid_input`. */
+    defaults(options?: {
+        agent?: string;
+    }): Promise<AnalyzeDefaults>;
     /**
      * The result document — the wire's TrialAnalysis, statuses and typed
      * failure included, for EVERY analysis (not only completed ones). The same
@@ -4583,10 +4597,10 @@ interface AnalysesClient {
      * one directory named as Harbor names the wrapper trial
      * (`analyze-<analyzed trial dir>__<7 chars>/`): config.json, lock.json,
      * result.json, trial.log, exception.txt (an infrastructure failure only),
-     * agent/claude-code.txt (Harbor's tee name for claude-code),
+     * agent/<Harbor's tee name for its agent> (claude-code.txt, codex.txt, …),
      * agent/stderr.log, agent/trace-parsed.jsonl, the captured home at its
-     * real names with agent/agent-home.json beside it and Harbor's copy at
-     * agent/sessions/, verifier/{test-stdout.txt,reward.txt,reward.json}
+     * real names with agent/agent-home.json beside it and Harbor's copy at its
+     * per-agent slot (agent/sessions/ for claude and codex), verifier/{test-stdout.txt,reward.txt,reward.json}
      * when the validator ruled (reward 1 = a valid analysis.json, 0 = it was
      * refused), and artifacts/manifest.json with artifacts/analysis.json (the
      * validated {summary, checks}) on a completed run — absent artifacts are
@@ -4609,8 +4623,8 @@ interface AnalysesClient {
 /**
  * Task-check configuration — Harbor's `harbor check` vocabulary (their
  * cli/analyze.py:84-148 check_command), the spec's `CheckConfigInput`. The
- * rubric-agent trio is the analyze door's, under the same rules
- * (`AnalyzeConfigInput` states them; refusals name `check.*`): `model_name`
+ * rubric-agent knobs are the analyze door's, under the same rules
+ * (`AnalyzeConfigInput` states them; refusals name `check.*`): `agent` (`-a/--agent`), `model_name`
  * (Harbor's check default is `claude-sonnet-4-6`; this platform's is the
  * analyzer's `openrouter/deepseek/deepseek-v4.1-flash` — one roster, one
  * default for both rubric
@@ -4628,14 +4642,15 @@ interface AnalysesClient {
  * (`-x/--exclude-task-name`), `n_tasks` (`-l/--n-tasks`) — applied in
  * checker.py's order (:132-138: include, exclude, then the cap) over the
  * sorted task directory names, Python fnmatch globs against the NAME.
- * Harbor's `-a/--agent`, `--job-name`, `-o/--jobs-dir`, `-k/--n-attempts`
- * and the local-runner kwargs are not on this surface; the contract records
- * each with its reason.
+ * Harbor's `-o/--jobs-dir`, `-k/--n-attempts` and the local-runner kwargs
+ * are not on this surface; the contract records each with its reason.
  */
 interface CheckConfigInput {
     /** A name for the check (Harbor's `--job-name`); omitted, the accept timestamp `YYYY-MM-DD__HH-MM-SS`. 1-120 characters. */
     name?: string;
-    /** Model the checker agent runs (Harbor's `-m/--model`); must be on the claude roster (`GET /api/meta`). */
+    /** The agent the checker runs on (Harbor's `-a/--agent`) — the analyze door's `agent`, same rule: a retired one is refused `agent_retired`. Omitted: claude. */
+    agent?: string;
+    /** Model the checker agent runs (Harbor's `-m/--model`) — the analyze door's `model_name`, same rule: omitted, the agent's default (`GET /api/meta` `analyze.agents[].default_model`); named, on the agent's roster. */
     model_name?: string;
     /** The rubric (Harbor's `-r/--rubric` file as its `{criteria}` object); default: the platform's check rubric (`checks().defaults()`). */
     rubric?: Rubric;
@@ -4740,9 +4755,11 @@ interface Check {
     name: string;
     status: CheckStatus;
     source: CheckSource;
+    /** The agent every task's checker ran on — named at create, or the default (claude). */
+    agent: string;
     model_name: string;
-    /** The effort every task's checker ran at — named at create, or the model's default, resolved at accept. */
-    reasoning_effort: string;
+    /** The effort every task's checker ran at — named at create, or the model's default on the agent, resolved at accept; null when the agent takes none. */
+    reasoning_effort: string | null;
     rubric: Rubric;
     /** The prompt template as stored; null = the platform's default check body. */
     prompt: string | null;
@@ -4770,33 +4787,38 @@ interface Check {
     finished_at: string | null;
 }
 /**
- * The policy an empty analyze config resolves to (GET /api/analyses/defaults):
- * each key the value `AnalyzeConfig` echoes for a job created with
- * `analyze: {}`, except `prompt`, which `AnalyzeConfig` serves as null and
- * this serves as the template text.
+ * The policy an empty analyze config resolves to (GET /api/analyses/defaults),
+ * or one naming only `agent` when `defaults({ agent })` names it: each key the
+ * value `AnalyzeConfig` echoes for a job created with that config, except
+ * `prompt`, which `AnalyzeConfig` serves as null and this serves as the
+ * template text.
  */
 interface AnalyzeDefaults {
+    agent: string;
+    /** The model an omitted `model_name` takes on `agent` (`GET /api/meta` `analyze.agents[].default_model`). */
     model_name: string;
     rubric: Rubric;
     /** The built-in analyze prompt template, unrendered — pass it as `prompt` to run the default body explicitly, or edit it from here. */
     prompt: string;
-    /** The effort the default model runs at when the config names none. */
-    reasoning_effort: string;
+    /** The effort `model_name` runs at on `agent` when the config names none; null if that agent takes none. */
+    reasoning_effort: string | null;
     sandbox_provider: EvalSandboxProvider;
 }
 /**
- * The policy an empty check config resolves to (GET /api/checks/defaults):
- * each key the value `Check` echoes for a check created with no config,
- * except `prompt`, which `Check` serves as null and this serves as the
- * template text.
+ * The policy an empty check config resolves to (GET /api/checks/defaults), or
+ * one naming only `agent` when `defaults({ agent })` names it: each key the
+ * value `Check` echoes for a check created with that config, except `prompt`,
+ * which `Check` serves as null and this serves as the template text.
  */
 interface CheckDefaults {
+    agent: string;
+    /** The model an omitted `model_name` takes on `agent` (`GET /api/meta` `analyze.agents[].default_model`). */
     model_name: string;
     rubric: Rubric;
     /** The built-in check prompt template, unrendered — pass it as `prompt` to run the default body explicitly, or edit it from here. */
     prompt: string;
-    /** The effort the default model runs at when the config names none. */
-    reasoning_effort: string;
+    /** The effort `model_name` runs at on `agent` when the config names none; null if that agent takes none. */
+    reasoning_effort: string | null;
     sandbox_provider: EvalSandboxProvider;
 }
 /** Options for checks().watch() */
@@ -4867,8 +4889,10 @@ interface ChecksClient {
     get(checkId: string): Promise<Check>;
     /** Every check you may read, newest first (cursor-paged); `{ scope, status, dataset }` narrow it. */
     list(options?: ListChecksOptions): CheckList;
-    /** The defaults a check runs under when its config names nothing (GET /api/checks/defaults): model, effort, provider, rubric and the unrendered prompt template. */
-    defaults(): Promise<CheckDefaults>;
+    /** The defaults a check runs under when its config names nothing (GET /api/checks/defaults): agent, model, effort, provider, rubric and the unrendered prompt template. `{ agent }` reads what that agent runs under — its default model and the effort it takes; an unknown name is refused `invalid_input`. */
+    defaults(options?: {
+        agent?: string;
+    }): Promise<CheckDefaults>;
     /** Poll a check until every task settled; resolves with the final Check. */
     watch(checkId: string, options?: WatchCheckOptions): Promise<Check>;
     /**
@@ -4901,9 +4925,9 @@ interface ChecksClient {
      * the TASK CHECK id (`Check.results[].id`). Each folder is Harbor's
      * TrialPaths for the checker's run: config.json, lock.json, result.json,
      * trial.log, exception.txt (an infrastructure failure only),
-     * agent/claude-code.txt, agent/stderr.log, agent/trace-parsed.jsonl, the
-     * captured home at its real names with agent/agent-home.json and Harbor's
-     * copy at agent/sessions/, verifier/{test-stdout.txt,reward.txt,
+     * agent/<its agent's tee name>, agent/stderr.log, agent/trace-parsed.jsonl, the
+     * captured home at its real names with agent/agent-home.json and Harbor's copy at its
+     * per-agent slot (agent/sessions/ for claude and codex), verifier/{test-stdout.txt,reward.txt,
      * reward.json} when the validator ruled (reward 1 = a valid
      * check-result.json, 0 = it was refused), artifacts/manifest.json and
      * artifacts/check-result.json (the validated flat checks) on a completed
@@ -5092,7 +5116,7 @@ interface OrgsClient {
  * that file existed. Adding a code means editing the spec, that file, this
  * list, and the Python pair.
  */
-declare const HOSTED_ERROR_CODES: readonly ["missing_authorization", "invalid_api_key", "read_only_key", "credential_service_unavailable", "rate_limited", "insufficient_credits", "quota_exceeded", "invalid_json", "invalid_input", "invalid_limit", "invalid_status", "invalid_visibility", "invalid_cursor", "invalid_after", "invalid_format", "invalid_ids", "invalid_multipart", "idempotency_key_reused", "dataset_not_found", "dataset_version_not_found", "dataset_name_taken", "dataset_in_use", "dataset_not_owned", "dataset_import_in_progress", "upstream_not_watchable", "no_active_version", "version_not_ready", "version_not_activatable", "unknown_task_names", "no_tasks", "task_not_found", "task_failed_to_build", "upload_session_not_found", "upload_offset_mismatch", "upload_chunk_digest_mismatch", "upload_incomplete", "upload_archive_digest_mismatch", "upload_session_failed", "too_many_concurrent_upload_chunks", "agent_not_found", "agent_name_taken", "agent_name_reserved", "agent_invalid_name", "agent_source_required", "agent_source_conflict", "agent_invalid_env", "agent_too_large", "agent_limit_reached", "skill_not_found", "skill_name_not_found", "skill_ref_invalid", "skill_unresolvable", "skill_invalid", "skill_in_use", "skill_too_large", "skill_limit_reached", "too_many_concurrent_skill_uploads", "secret_not_found", "secret_ambiguous", "secret_brokered_unsupported", "secret_exists", "secret_not_attached", "agent_version_not_found", "agent_version_unresolvable", "agent_kwarg_unsupported", "agent_config_unsupported", "agent_config_key_refused", "agent_preset_unsupported", "provider_unsupported", "job_not_found", "job_not_terminal", "no_failed_trials", "trial_not_found", "trial_not_settled", "concurrent_update", "regrade_source_ineligible", "no_regradable_trials", "invalid_rubric", "analysis_already_running", "analysis_not_found", "analysis_not_terminal", "check_not_found", "check_not_terminal", "no_checkable_tasks", "too_many_concurrent_check_uploads", "no_analyzable_trials", "session_not_found", "not_a_job_dir", "invalid_trial", "trial_too_large", "upload_too_large", "job_uploaded", "job_already_uploaded", "too_many_concurrent_job_uploads", "job_import_not_found", "import_not_found", "import_too_large", "too_many_concurrent_imports", "invalid_archive", "unpinned_git_ref", "hub_package_not_found", "hub_unreachable", "package_not_retained", "package_corrupt", "package_missing", "too_many_concurrent_package_downloads", "org_not_found", "org_slug_taken", "org_forbidden", "org_personal_immutable", "org_last_owner", "org_in_use", "org_member_not_found", "invite_not_found", "invite_invalid", "not_found", "not_captured", "filesystem_state", "feature_unsupported", "provider_unreachable", "task_package_not_retained", "internal_error"];
+declare const HOSTED_ERROR_CODES: readonly ["missing_authorization", "invalid_api_key", "read_only_key", "credential_service_unavailable", "rate_limited", "insufficient_credits", "quota_exceeded", "invalid_json", "invalid_input", "invalid_limit", "invalid_status", "invalid_visibility", "invalid_cursor", "invalid_after", "invalid_format", "invalid_ids", "invalid_multipart", "idempotency_key_reused", "dataset_not_found", "dataset_version_not_found", "dataset_name_taken", "dataset_in_use", "dataset_not_owned", "dataset_import_in_progress", "upstream_not_watchable", "no_active_version", "version_not_ready", "version_not_activatable", "unknown_task_names", "no_tasks", "task_not_found", "task_failed_to_build", "upload_session_not_found", "upload_offset_mismatch", "upload_chunk_digest_mismatch", "upload_incomplete", "upload_archive_digest_mismatch", "upload_session_failed", "too_many_concurrent_upload_chunks", "agent_not_found", "agent_name_taken", "agent_name_reserved", "agent_invalid_name", "agent_source_required", "agent_source_conflict", "agent_invalid_env", "agent_too_large", "agent_limit_reached", "skill_not_found", "skill_name_not_found", "skill_ref_invalid", "skill_unresolvable", "skill_invalid", "skill_in_use", "skill_too_large", "skill_limit_reached", "too_many_concurrent_skill_uploads", "secret_not_found", "secret_ambiguous", "secret_brokered_unsupported", "secret_exists", "secret_not_attached", "agent_version_not_found", "agent_version_unresolvable", "agent_kwarg_unsupported", "agent_config_unsupported", "agent_config_key_refused", "agent_preset_unsupported", "agent_retired", "provider_unsupported", "job_not_found", "job_not_terminal", "no_failed_trials", "trial_not_found", "trial_not_settled", "concurrent_update", "regrade_source_ineligible", "no_regradable_trials", "invalid_rubric", "analysis_already_running", "analysis_not_found", "analysis_not_terminal", "check_not_found", "check_not_terminal", "no_checkable_tasks", "too_many_concurrent_check_uploads", "no_analyzable_trials", "session_not_found", "not_a_job_dir", "invalid_trial", "trial_too_large", "upload_too_large", "job_uploaded", "job_already_uploaded", "too_many_concurrent_job_uploads", "job_import_not_found", "import_not_found", "import_too_large", "too_many_concurrent_imports", "invalid_archive", "unpinned_git_ref", "hub_package_not_found", "hub_unreachable", "package_not_retained", "package_corrupt", "package_missing", "too_many_concurrent_package_downloads", "org_not_found", "org_slug_taken", "org_forbidden", "org_personal_immutable", "org_last_owner", "org_in_use", "org_member_not_found", "invite_not_found", "invite_invalid", "not_found", "not_captured", "filesystem_state", "feature_unsupported", "provider_unreachable", "task_package_not_retained", "internal_error"];
 /** One of the API's stable error codes. */
 type HostedErrorCode = (typeof HOSTED_ERROR_CODES)[number];
 /** True when `value` is a code this SDK version knows about (narrowing guard). */
@@ -5117,6 +5141,11 @@ interface AgentModelOption {
     model_id: string;
     description: string | null;
 }
+/** One retired built-in agent and the agent to use instead — the row `agent_retired` details carry too. */
+interface RetiredAgent {
+    agent: string;
+    replaced_by: string;
+}
 /** One built-in agent's declared capabilities. */
 interface AgentCapability {
     name: string;
@@ -5228,8 +5257,15 @@ interface ManagedProviderCapability {
  */
 interface CapabilityDocument {
     schema_version: number;
-    /** Built-in agents and their declared capabilities. */
+    /** Built-in agents a new job may name, and their declared capabilities. */
     agents: AgentCapability[];
+    /**
+     * Retired built-in agents, left out of `agents`: a new job, resume, retry,
+     * analysis or check naming one is refused `agent_retired`, while its records
+     * stay readable.
+     * Absent on servers predating the field.
+     */
+    retired_agents?: RetiredAgent[];
     /** Rules a bring-your-own agent registration must satisfy. */
     agent_registration: {
         name_pattern: string;
@@ -5267,19 +5303,24 @@ interface CapabilityDocument {
         dataset_version: StatusVocabulary;
     };
     /**
-     * The trace analyzer's roster and defaults: the model an omitted
-     * `analyze.model_name` takes, the efforts `analyze.reasoning_effort`
-     * accepts (the claude harness's — the analyzer IS that harness), and
-     * every roster model with the effort an omitted `reasoning_effort` takes
-     * for it. Absent on servers predating the field.
+     * The rubric agents' roster and defaults, the analyzer's and the checker's alike, so a client
+     * knows what "omitted" meant. Absent on servers predating the field.
      */
     analyze?: {
+        default_agent: string;
+        /** The platform's pick — what an omitted `model_name` takes on every agent whose roster carries it. */
         default_model: string;
-        reasoning_efforts: string[];
-        models: {
-            alias: string;
-            model_id: string;
-            default_reasoning_effort: string;
+        agents: {
+            name: string;
+            /** What an omitted `model_name` takes on this agent; one of its `models[].alias`. */
+            default_model: string;
+            /** Empty for an agent that takes no effort. */
+            reasoning_efforts: string[];
+            models: {
+                alias: string;
+                model_id: string;
+                default_reasoning_effort: string | null;
+            }[];
         }[];
     };
     limits: {
@@ -5375,4 +5416,4 @@ interface CapabilityDocument {
     error_codes: string[];
 }
 
-export { type AnalysisEvidence as $, type AgentInput as A, type AgentArm as B, type Check as C, type DatasetVersion as D, type AgentArmInput as E, type AgentCapability as F, type GatewayUsageEvent as G, type HostedErrorCode as H, type AgentDatasetStats as I, type JobCreate as J, type AgentEffortSupport as K, type AgentInfo as L, type AgentList as M, type AgentModelOption as N, type OrgsClient as O, type PublishDatasetInput as P, type AgentPage as Q, type Rubric as R, type SkillsClient as S, type Trial as T, type UsageReading as U, type AgentResult as V, type AgentSource as W, type AgentSourceInput as X, type AgentUpsertInput as Y, type AnalysisArtifactStream as Z, type AnalysisCheck as _, type TrialAnalysis as a, type FilesystemSource as a$, type AnalysisFailure as a0, type AnalysisList as a1, type AnalysisPage as a2, type AnalysisStatus as a3, type AnalysisTranscript as a4, type AnalysisTranscriptOptions as a5, type AnalyzeConfig as a6, type AnalyzeConfigInput as a7, type AnalyzeDefaults as a8, type ApiKey as a9, type DatasetVersionArchiveSource as aA, type DatasetVersionArchiveUrlSource as aB, type DatasetVersionGitSource as aC, type DatasetVersionHubSource as aD, type DatasetVersionSource as aE, type DatasetVersionState as aF, type DownloadDatasetOptions as aG, type DownloadJobOptions as aH, EVAL_SANDBOX_PROVIDERS as aI, type Agent as aJ, type ModelInfo as aK, type EvalSandboxProvider as aL, type StepResult as aM, type ExceptionInfo as aN, type FilesystemArchiveOptions as aO, type FilesystemBox as aP, type FilesystemCapture as aQ, type FilesystemChange as aR, type FilesystemChanges as aS, type FilesystemChangesOptions as aT, type FilesystemEntry as aU, type FilesystemListOptions as aV, type FilesystemListing as aW, type FilesystemReadOptions as aX, type FilesystemSearchHit as aY, type FilesystemSearchOptions as aZ, type FilesystemSearchResult as a_, type AttemptPhase as aa, type AuthStatus as ab, type Awaitable as ac, CHECK_STATUSES as ad, type CheckDefaults as ae, type CheckRow as af, type CheckRowList as ag, type CheckRowPage as ah, type CheckTaskTally as ai, type CompareCell as aj, type CompareCoverage as ak, type CompareJobAggregate as al, type CompareResponse as am, type CompareTaskRow as an, type Dataset as ao, type DatasetImportFailure as ap, type DatasetImportList as aq, type DatasetImportPage as ar, type DatasetImportStatus as as, type DatasetList as at, type DatasetPage as au, type DatasetPatch as av, type DatasetPreflight as aw, type DatasetRef as ax, type DatasetSelector as ay, type DatasetSource as az, type DatasetImport as b, type OrgQuota as b$, type FilesystemState as b0, type FilesystemStatus as b1, type FilesystemStreamEvent as b2, type FilesystemStreamOptions as b3, type FilesystemWatchResult as b4, GATEWAY_TRACE_SEQ_BASE as b5, type GatewayUsage as b6, type GetDatasetOptions as b7, type GrepJobOptions as b8, HOSTED_ERROR_CODES as b9, type JobPage as bA, type JobShareEmail as bB, type JobShareLink as bC, type JobStats as bD, type JobStatus as bE, type JobTaskLink as bF, type JobTaskRollup as bG, type JobTaskRollupList as bH, type JobTaskRollupPage as bI, type JobViewer as bJ, type JobWatch as bK, type JudgeResult as bL, type ListAgentsOptions as bM, type ListAnalysesOptions as bN, type ListDatasetsOptions as bO, type ListImportsOptions as bP, type ListJobImportsOptions as bQ, type ListJobTasksOptions as bR, type ListJobsOptions as bS, type ListSkillsOptions as bT, type ListTrialFilesOptions as bU, type ListTrialsOptions as bV, type ManagedProviderCapability as bW, type OrgInvite as bX, type OrgInviteCreated as bY, type OrgJoined as bZ, type OrgMember as b_, type ImportPhase as ba, type ImportPhaseProgress as bb, type ImportWarning as bc, type InfraFailureSignature as bd, JOB_LIST_KINDS as be, JOB_LIST_SCOPES as bf, type JobAnalysisStats as bg, type JobDeleteResult as bh, type JobFailure as bi, type JobGrepGroup as bj, type JobGrepPage as bk, type JobImport as bl, type JobImportFailure as bm, type JobImportList as bn, type JobImportPage as bo, type JobImportPhaseName as bp, type JobImportPhaseProgress as bq, type JobImportProgress as br, type JobImportSkippedTrial as bs, type JobImportSource as bt, type JobList as bu, type JobListItem as bv, type JobListItemList as bw, type JobListItemPage as bx, type JobListKind as by, type JobListScope as bz, type JobEvent as c, type UploadProvenance as c$, type OrgRole as c0, type OrgUsage as c1, type Organization as c2, type OrganizationDetail as c3, type Page as c4, type PageOptions as c5, type PassAtKGroup as c6, type PassAtKPoint as c7, type PreflightDatasetInput as c8, type PreflightDeferredCheck as c9, TASK_LINK_REASONS as cA, TRIAL_ARTIFACT_STREAMS as cB, TRIAL_STATUSES as cC, type Task as cD, type TaskCheckTranscript as cE, type TaskLinkReason as cF, type TaskLinkedBy as cG, type TaskNote as cH, type TaskPackageFiles as cI, type TaskPackageFilesystemStatus as cJ, type TaskProviderVerdict as cK, type TimingInfo as cL, type TraceEventPage as cM, type TraceOptions as cN, type TrialArtifactStream as cO, type TrialCounts as cP, type TrialFile as cQ, type TrialFilePage as cR, type TrialFileRange as cS, type TrialGpuCost as cT, type TrialList as cU, type TrialPage as cV, type TrialStatus as cW, type TrialStatusTally as cX, type TrialTaskLink as cY, type TrialUploadProvenance as cZ, type UploadJobOptions as c_, type PreflightManifestVerdict as ca, type PreflightTaskVerdict as cb, type ProviderCapability as cc, type PublishDatasetOptions as cd, type RegradeRequest as ce, type ResumeRequest as cf, type RetryRequest as cg, type RubricCriterion as ch, type RunFilesystem as ci, SANDBOX_LOG_STREAMS as cj, type SandboxLogEvent as ck, type SandboxLogLine as cl, type SandboxLogLines as cm, type SandboxLogOptions as cn, type SandboxLogStream as co, type SandboxProcs as cp, type SkillLock as cq, type SkillUpload as cr, type SkillUploadList as cs, type SkillUploadPage as ct, type SourceJob as cu, type SpendSource as cv, type StartJobOptions as cw, type StatusVocabulary as cx, type StopResponse as cy, TASK_LINKED_BY as cz, type DatasetFailedTask as d, type UploadSkillOptions as d0, type UpstreamStatus as d1, type VerifierEnvironmentMode as d2, type VerifierResult as d3, type WatchAnalysisOptions as d4, type WatchImportOptions as d5, type WatchJobImportOptions as d6, type WatchJobOptions as d7, gatewayUsageOf as d8, isHostedErrorCode as d9, passAtK as da, type DatasetImportProgress as e, type JobSecretInline as f, type JobSecretRef as g, type TraceEvent as h, type JobVisibility as i, type JobShareRequest as j, type JobShares as k, type Job as l, type TaskCheck as m, type DatasetsClient as n, type AgentsClient as o, type JobsClient as p, type TrialsClient as q, type AnalysesClient as r, type ChecksClient as s, type CapabilityDocument as t, type HostedClientConfig as u, type AuthClient as v, AGENT_EFFORT_SUPPORT_VALUES as w, ANALYSIS_ARTIFACT_STREAMS as x, ANALYSIS_STATUSES as y, type ActiveDataset as z };
+export { type AnalysisEvidence as $, type AgentInput as A, type AgentArm as B, type Check as C, type DatasetVersion as D, type AgentArmInput as E, type AgentCapability as F, type GatewayUsageEvent as G, type HostedErrorCode as H, type AgentDatasetStats as I, type JobCreate as J, type AgentEffortSupport as K, type AgentInfo as L, type AgentList as M, type AgentModelOption as N, type OrgsClient as O, type PublishDatasetInput as P, type AgentPage as Q, type Rubric as R, type SkillsClient as S, type Trial as T, type UsageReading as U, type AgentResult as V, type AgentSource as W, type AgentSourceInput as X, type AgentUpsertInput as Y, type AnalysisArtifactStream as Z, type AnalysisCheck as _, type TrialAnalysis as a, type FilesystemSource as a$, type AnalysisFailure as a0, type AnalysisList as a1, type AnalysisPage as a2, type AnalysisStatus as a3, type AnalysisTranscript as a4, type AnalysisTranscriptOptions as a5, type AnalyzeConfig as a6, type AnalyzeConfigInput as a7, type AnalyzeDefaults as a8, type ApiKey as a9, type DatasetVersionArchiveSource as aA, type DatasetVersionArchiveUrlSource as aB, type DatasetVersionGitSource as aC, type DatasetVersionHubSource as aD, type DatasetVersionSource as aE, type DatasetVersionState as aF, type DownloadDatasetOptions as aG, type DownloadJobOptions as aH, EVAL_SANDBOX_PROVIDERS as aI, type Agent as aJ, type ModelInfo as aK, type EvalSandboxProvider as aL, type StepResult as aM, type ExceptionInfo as aN, type FilesystemArchiveOptions as aO, type FilesystemBox as aP, type FilesystemCapture as aQ, type FilesystemChange as aR, type FilesystemChanges as aS, type FilesystemChangesOptions as aT, type FilesystemEntry as aU, type FilesystemListOptions as aV, type FilesystemListing as aW, type FilesystemReadOptions as aX, type FilesystemSearchHit as aY, type FilesystemSearchOptions as aZ, type FilesystemSearchResult as a_, type AttemptPhase as aa, type AuthStatus as ab, type Awaitable as ac, CHECK_STATUSES as ad, type CheckDefaults as ae, type CheckRow as af, type CheckRowList as ag, type CheckRowPage as ah, type CheckTaskTally as ai, type CompareCell as aj, type CompareCoverage as ak, type CompareJobAggregate as al, type CompareResponse as am, type CompareTaskRow as an, type Dataset as ao, type DatasetImportFailure as ap, type DatasetImportList as aq, type DatasetImportPage as ar, type DatasetImportStatus as as, type DatasetList as at, type DatasetPage as au, type DatasetPatch as av, type DatasetPreflight as aw, type DatasetRef as ax, type DatasetSelector as ay, type DatasetSource as az, type DatasetImport as b, type OrgQuota as b$, type FilesystemState as b0, type FilesystemStatus as b1, type FilesystemStreamEvent as b2, type FilesystemStreamOptions as b3, type FilesystemWatchResult as b4, GATEWAY_TRACE_SEQ_BASE as b5, type GatewayUsage as b6, type GetDatasetOptions as b7, type GrepJobOptions as b8, HOSTED_ERROR_CODES as b9, type JobPage as bA, type JobShareEmail as bB, type JobShareLink as bC, type JobStats as bD, type JobStatus as bE, type JobTaskLink as bF, type JobTaskRollup as bG, type JobTaskRollupList as bH, type JobTaskRollupPage as bI, type JobViewer as bJ, type JobWatch as bK, type JudgeResult as bL, type ListAgentsOptions as bM, type ListAnalysesOptions as bN, type ListDatasetsOptions as bO, type ListImportsOptions as bP, type ListJobImportsOptions as bQ, type ListJobTasksOptions as bR, type ListJobsOptions as bS, type ListSkillsOptions as bT, type ListTrialFilesOptions as bU, type ListTrialsOptions as bV, type ManagedProviderCapability as bW, type OrgInvite as bX, type OrgInviteCreated as bY, type OrgJoined as bZ, type OrgMember as b_, type ImportPhase as ba, type ImportPhaseProgress as bb, type ImportWarning as bc, type InfraFailureSignature as bd, JOB_LIST_KINDS as be, JOB_LIST_SCOPES as bf, type JobAnalysisStats as bg, type JobDeleteResult as bh, type JobFailure as bi, type JobGrepGroup as bj, type JobGrepPage as bk, type JobImport as bl, type JobImportFailure as bm, type JobImportList as bn, type JobImportPage as bo, type JobImportPhaseName as bp, type JobImportPhaseProgress as bq, type JobImportProgress as br, type JobImportSkippedTrial as bs, type JobImportSource as bt, type JobList as bu, type JobListItem as bv, type JobListItemList as bw, type JobListItemPage as bx, type JobListKind as by, type JobListScope as bz, type JobEvent as c, type UploadJobOptions as c$, type OrgRole as c0, type OrgUsage as c1, type Organization as c2, type OrganizationDetail as c3, type Page as c4, type PageOptions as c5, type PassAtKGroup as c6, type PassAtKPoint as c7, type PreflightDatasetInput as c8, type PreflightDeferredCheck as c9, TASK_LINKED_BY as cA, TASK_LINK_REASONS as cB, TRIAL_ARTIFACT_STREAMS as cC, TRIAL_STATUSES as cD, type Task as cE, type TaskCheckTranscript as cF, type TaskLinkReason as cG, type TaskLinkedBy as cH, type TaskNote as cI, type TaskPackageFiles as cJ, type TaskPackageFilesystemStatus as cK, type TaskProviderVerdict as cL, type TimingInfo as cM, type TraceEventPage as cN, type TraceOptions as cO, type TrialArtifactStream as cP, type TrialCounts as cQ, type TrialFile as cR, type TrialFilePage as cS, type TrialFileRange as cT, type TrialGpuCost as cU, type TrialList as cV, type TrialPage as cW, type TrialStatus as cX, type TrialStatusTally as cY, type TrialTaskLink as cZ, type TrialUploadProvenance as c_, type PreflightManifestVerdict as ca, type PreflightTaskVerdict as cb, type ProviderCapability as cc, type PublishDatasetOptions as cd, type RegradeRequest as ce, type ResumeRequest as cf, type RetiredAgent as cg, type RetryRequest as ch, type RubricCriterion as ci, type RunFilesystem as cj, SANDBOX_LOG_STREAMS as ck, type SandboxLogEvent as cl, type SandboxLogLine as cm, type SandboxLogLines as cn, type SandboxLogOptions as co, type SandboxLogStream as cp, type SandboxProcs as cq, type SkillLock as cr, type SkillUpload as cs, type SkillUploadList as ct, type SkillUploadPage as cu, type SourceJob as cv, type SpendSource as cw, type StartJobOptions as cx, type StatusVocabulary as cy, type StopResponse as cz, type DatasetFailedTask as d, type UploadProvenance as d0, type UploadSkillOptions as d1, type UpstreamStatus as d2, type VerifierEnvironmentMode as d3, type VerifierResult as d4, type WatchAnalysisOptions as d5, type WatchImportOptions as d6, type WatchJobImportOptions as d7, type WatchJobOptions as d8, gatewayUsageOf as d9, isHostedErrorCode as da, passAtK as db, type DatasetImportProgress as e, type JobSecretInline as f, type JobSecretRef as g, type TraceEvent as h, type JobVisibility as i, type JobShareRequest as j, type JobShares as k, type Job as l, type TaskCheck as m, type DatasetsClient as n, type AgentsClient as o, type JobsClient as p, type TrialsClient as q, type AnalysesClient as r, type ChecksClient as s, type CapabilityDocument as t, type HostedClientConfig as u, type AuthClient as v, AGENT_EFFORT_SUPPORT_VALUES as w, ANALYSIS_ARTIFACT_STREAMS as x, ANALYSIS_STATUSES as y, type ActiveDataset as z };