vigiles 19.0.1 → 21.0.0

package/dist/services.js

@@ -3,7 +3,7 @@
  * vigiles — R3 disposable-service tier (⚠️ EXPERIMENTAL / UNSTABLE).
  *
  * ─────────────────────────────────────────────────────────────────────────────
- *   EXPERIMENTAL: this surface is a DRAFT. Import it from `vigiles/experimental`,
+ *   EXPERIMENTAL: this surface is a DRAFT. Import it from `vigiles`,
  *   NOT from a stable subpath. It is NOT covered by the stability guarantee and
  *   may change shape or be removed WITHOUT a major-version bump. Do not build a
  *   production workflow on it yet. See docs/measuring-skills.md § Experimental.
@@ -45,7 +45,7 @@
  * side-effect-free).
  *
  * @experimental
- * @module vigiles/experimental (services)
+ * @module vigiles (services)
  */
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.experimental_startServices = experimental_startServices;

package/dist/test.d.ts

@@ -56,6 +56,7 @@ export type { RunScriptOptions, ScriptRunResult } from "./run-script.js";
 export { runHook, parseHookOutput, decideHook, propertyHook, fileToolEvents, egressRoutes, } from "./run-hook.js";
 export type { HookRunResult, RunHookOptions, HookInput, HookOutput, HookPropertyResult, FileToolEventOptions, } from "./run-hook.js";
 export * from "./harness-assert.js";
+export { experimental_emitTool, experimental_parseEmitted, experimental_assertEmittedOk, type EmitFieldSchema, type EmitObjectSchema, type EmitPropertySchema, type EmitTrackSchema, type EmitToolDefinition, type ExperimentalEmitTool, } from "./experimental-emit.js";
 export { loadHook } from "./load-hook.js";
 export { evalChecks, assertChecks, tool, toolWith, notTool, onlyTools, skill, output, hookFired, received, turns, wrote, didNotWrite, subagent, blocked, allowed, mcp, cost, latency, tokens, inputTokens, outputTokens, cacheTokens, } from "./check.js";
 export type { ArgMatcher, Check, CheckJSON, CheckResult, JudgeFn, } from "./check.js";
@@ -74,4 +75,6 @@ export { defineEval } from "./eval-define.js";
 export type { EvalDefinition, EvalDefinitionInput, EvalHooks, EvalKind, EvalMeasurements, EvalReports, SelectionMatrixSpec, } from "./eval-define.js";
 export { assertRates, assertPromptDiversity, checkPromptDiversity, checkReportToJUnit, formatCheckReport, formatEvalReport, formatTriggerRateReport, parseClaudeRun, stubSkillBody, } from "./eval.js";
 export type { EvalArm, EvalDriver, EvalSpec, EvalReport, EvalUsage, MeasureSpec, ArmsMeasureSpec, ArmReport, ArmUsage, ArmsCheckReport, CheckRate, CheckReport, MetricStat, Metrics, ModelOutputParser, ParsedModelRun, PromptDiversityIssue, PromptTriggerStat, RunContext, RunOut, SelectionTrialResult, TriggerRateReport, TriggerRateSpec, AgentRunArgs, AgentRunner, } from "./eval.js";
+export { experimental_startServices, experimental_withServices, type ServiceSpec, type ServiceReady, type ServiceReset, type ServiceHandle, type ServiceSession, type ContainerRuntime, } from "./services.js";
+export { experimental_dockerRuntime, experimental_makeDockerRuntime, type DockerExec, type NetProbe, } from "./services-docker.js";
 //# sourceMappingURL=test.d.ts.map

package/dist/test.js

@@ -66,8 +66,8 @@ var __exportStar = (this && this.__exportStar) || function(m, exports) {
     for (var p in m) if (p !== "default" && !Object.prototype.hasOwnProperty.call(exports, p)) __createBinding(exports, m, p);
 };
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.mustNotInclude = exports.mustInclude = exports.commandsIn = exports.sandboxAvailable = exports.specTrusted = exports.decideSandbox = exports.parseHooks = exports.parseOutput = exports.parseResultEvent = exports.parseSubagents = exports.parseToolCalls = exports.runHarness = exports.runHarnessTest = exports.formatGuardrailReport = exports.assertBlocksDisasters = exports.unblockedDisasters = exports.verifyGuardrail = exports.DISASTER_CATALOG = exports.cacheTokens = exports.outputTokens = exports.inputTokens = exports.tokens = exports.latency = exports.cost = exports.mcp = exports.allowed = exports.blocked = exports.subagent = exports.didNotWrite = exports.wrote = exports.turns = exports.received = exports.hookFired = exports.output = exports.skill = exports.onlyTools = exports.notTool = exports.toolWith = exports.tool = exports.assertChecks = exports.evalChecks = exports.loadHook = exports.egressRoutes = exports.fileToolEvents = exports.propertyHook = exports.decideHook = exports.parseHookOutput = exports.runHook = exports.runScript = exports.recordCheck = void 0;
-exports.stubSkillBody = exports.parseClaudeRun = exports.formatTriggerRateReport = exports.formatEvalReport = exports.formatCheckReport = exports.checkReportToJUnit = exports.checkPromptDiversity = exports.assertPromptDiversity = exports.assertRates = exports.defineEval = exports.formatContainment = exports.compareContainment = exports.skillContract = void 0;
+exports.sandboxAvailable = exports.specTrusted = exports.decideSandbox = exports.parseHooks = exports.parseOutput = exports.parseResultEvent = exports.parseSubagents = exports.parseToolCalls = exports.runHarness = exports.runHarnessTest = exports.formatGuardrailReport = exports.assertBlocksDisasters = exports.unblockedDisasters = exports.verifyGuardrail = exports.DISASTER_CATALOG = exports.cacheTokens = exports.outputTokens = exports.inputTokens = exports.tokens = exports.latency = exports.cost = exports.mcp = exports.allowed = exports.blocked = exports.subagent = exports.didNotWrite = exports.wrote = exports.turns = exports.received = exports.hookFired = exports.output = exports.skill = exports.onlyTools = exports.notTool = exports.toolWith = exports.tool = exports.assertChecks = exports.evalChecks = exports.loadHook = exports.experimental_assertEmittedOk = exports.experimental_parseEmitted = exports.experimental_emitTool = exports.egressRoutes = exports.fileToolEvents = exports.propertyHook = exports.decideHook = exports.parseHookOutput = exports.runHook = exports.runScript = exports.recordCheck = void 0;
+exports.experimental_makeDockerRuntime = exports.experimental_dockerRuntime = exports.experimental_withServices = exports.experimental_startServices = exports.stubSkillBody = exports.parseClaudeRun = exports.formatTriggerRateReport = exports.formatEvalReport = exports.formatCheckReport = exports.checkReportToJUnit = exports.checkPromptDiversity = exports.assertPromptDiversity = exports.assertRates = exports.defineEval = exports.formatContainment = exports.compareContainment = exports.skillContract = exports.mustNotInclude = exports.mustInclude = exports.commandsIn = void 0;
 // --- reporting: how much did this script actually do? ---
 // `vigiles test` can otherwise see only an exit code, so a file that runs NOTHING
 // prints the same `✓` as one that ran and passed (measured 2026-08-08 on a file
@@ -92,6 +92,19 @@ Object.defineProperty(exports, "fileToolEvents", { enumerable: true, get: functi
 Object.defineProperty(exports, "egressRoutes", { enumerable: true, get: function () { return run_hook_js_1.egressRoutes; } });
 // --- assertions and the eval-RESULT analysis helpers (free; see cost #1 above) ---
 __exportStar(require("./harness-assert.js"), exports);
+// --- the EMIT delivery for a typed result (⚠️ EXPERIMENTAL) ---
+// Moved here 2026-08-21 from the `vigiles/experimental` subpath, which is gone
+// (the `experimental_` prefix already marks every call site; the subpath only
+// marked the import line). It belongs BESIDE the assertions above rather than in
+// a drawer of its own: `experimental_parseEmitted` / `experimental_assertEmittedOk`
+// are the emit-channel twins of `parseAgentResult` / `assertAgentOk`, and a reader
+// comparing the two delivery shapes should not have to change doors to see both.
+// Read src/experimental-emit.ts's header before using it — it lists, by number,
+// what is unproven and what would have to be true to drop the prefix.
+var experimental_emit_js_1 = require("./experimental-emit.js");
+Object.defineProperty(exports, "experimental_emitTool", { enumerable: true, get: function () { return experimental_emit_js_1.experimental_emitTool; } });
+Object.defineProperty(exports, "experimental_parseEmitted", { enumerable: true, get: function () { return experimental_emit_js_1.experimental_parseEmitted; } });
+Object.defineProperty(exports, "experimental_assertEmittedOk", { enumerable: true, get: function () { return experimental_emit_js_1.experimental_assertEmittedOk; } });
 // The compiled-hook LOADER. The in-process assertions above take the hook
 // OBJECT, but a `.harness.mjs` test only has its PATH — without this the
 // intended in-process test path is unreachable from the file format
@@ -196,4 +209,31 @@ Object.defineProperty(exports, "formatEvalReport", { enumerable: true, get: func
 Object.defineProperty(exports, "formatTriggerRateReport", { enumerable: true, get: function () { return eval_js_1.formatTriggerRateReport; } });
 Object.defineProperty(exports, "parseClaudeRun", { enumerable: true, get: function () { return eval_js_1.parseClaudeRun; } });
 Object.defineProperty(exports, "stubSkillBody", { enumerable: true, get: function () { return eval_js_1.stubSkillBody; } });
+// --- R3: the disposable-service tier (⚠️ EXPERIMENTAL, and it has REAL EFFECTS) ---
+// Moved off the `vigiles/experimental` subpath 2026-08-21, when that subpath was
+// deleted: it marked the IMPORT LINE, which is out of view by the time anyone
+// reads the call, while `experimental_` marks every call site.
+//
+// 🔴 IT LANDED ON `vigiles/eval` FIRST, AND THE EXPORT-PREFIX GATE REJECTED IT —
+// correctly, and the reasoning is worth keeping. `./eval` has a naming contract:
+// every runtime export there is `paid_*`, meaning THIS CALL CAN BILL YOU. R3 felt
+// like it belonged because the docs frame it as measuring skills for real. But
+// `experimental_startServices` takes a `ContainerRuntime` and starts containers;
+// it calls no model. Naming it `paid_experimental_startServices` to satisfy the
+// gate would have made `paid_` mean "expensive in some sense", which is exactly
+// the erosion that turns a prefix back into decoration. The model spend is in
+// `paid_measureArms`, which is already over there. So R3 is free-of-model-cost
+// and belongs on this door — where "free" keeps its narrow meaning.
+//
+// ⚠️ FREE OF MODEL COST IS NOT FREE OF CONSEQUENCE, and this is the one place the
+// root barrel carries something with real side effects. The disposable container
+// is the ONLY isolation vigiles provides — it does NOT confine the skill's
+// filesystem or network. Run it somewhere disposable with no production access
+// and keep real credentials out of the run. See the SAFETY note in src/services.ts.
+var services_js_1 = require("./services.js");
+Object.defineProperty(exports, "experimental_startServices", { enumerable: true, get: function () { return services_js_1.experimental_startServices; } });
+Object.defineProperty(exports, "experimental_withServices", { enumerable: true, get: function () { return services_js_1.experimental_withServices; } });
+var services_docker_js_1 = require("./services-docker.js");
+Object.defineProperty(exports, "experimental_dockerRuntime", { enumerable: true, get: function () { return services_docker_js_1.experimental_dockerRuntime; } });
+Object.defineProperty(exports, "experimental_makeDockerRuntime", { enumerable: true, get: function () { return services_docker_js_1.experimental_makeDockerRuntime; } });
 //# sourceMappingURL=test.js.map

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vigiles",
-  "version": "19.0.1",
+  "version": "21.0.0",
   "description": "Audit, test and measure the harness your AI agent runs on — grade your CLAUDE.md / AGENTS.md, skills, subagents and hooks, run them against a scripted model, and measure whether they actually fire.",
   "keywords": [
     "claude-code",
@@ -46,7 +46,6 @@
     "./claude-code": "./dist/claude-code.js",
     "./codex": "./dist/codex.js",
     "./adapter": "./dist/adapter.js",
-    "./experimental": "./dist/experimental.js",
     "./vitest": {
       "types": "./dist/vitest.d.mts",
       "default": "./dist/vitest.mjs"
@@ -92,7 +91,7 @@
     "check": "node scripts/check.mjs",
     "docs:check": "npm run build && node scripts/check-doc-imports.mjs . docs README.md",
     "exports:check": "npm run build && node scripts/check-export-prefixes.mjs .",
-    "experimental:check": "npm run api:check && node scripts/check-experimental-naming.mjs",
+    "internal:check": "npm run api:check && node scripts/check-internal-tag.mjs",
     "docs:api": "typedoc"
   },
   "devDependencies": {

package/dist/experimental.d.ts

@@ -1,34 +0,0 @@
-/**
- * `vigiles/experimental` — ⚠️ EXPERIMENTAL, UNSTABLE public surface.
- *
- * Everything re-exported here is a DRAFT. It is deliberately quarantined behind
- * the `experimental` subpath (and the `experimental_` name prefix on runtime
- * exports) so the import itself signals the risk at the call site:
- *
- *   import { experimental_startServices } from "vigiles/experimental";
- *
- * NOT covered by the stability guarantee (STABILITY.md): the shape may change or
- * be removed WITHOUT a major-version bump. Do not depend on it in production.
- *
- * Current contents:
- *  - the R3 disposable-service tier (real side-effect testing; see
- *    docs/measuring-skills.md § Experimental and src/services.ts);
- *  - the EMIT delivery for a typed result (`src/experimental-emit.ts`) — a skill
- *    that CALLS a tool with its outcome instead of ending its turn with a fenced
- *    block, which is how an UNFORKED skill can carry an `OutputContract` at all.
- *    Read that module's header before using it: it lists, by number, what is
- *    unproven and what would have to be true to drop the prefix.
- *
- * ⚠️ SAFETY: R3 runs a model-driven skill FOR REAL. The disposable container is
- * the ONLY isolation vigiles provides — it does not confine the skill's filesystem
- * or network. Run it in a disposable environment with NO production access and
- * keep real credentials out of the run. See the SAFETY note in src/services.ts and
- * docs/measuring-skills.md § Experimental.
- *
- * @experimental
- * @module vigiles/experimental
- */
-export { experimental_startServices, experimental_withServices, type ServiceSpec, type ServiceReady, type ServiceReset, type ServiceHandle, type ServiceSession, type ContainerRuntime, } from "./services.js";
-export { experimental_dockerRuntime, experimental_makeDockerRuntime, type DockerExec, type NetProbe, } from "./services-docker.js";
-export { experimental_emitTool, experimental_parseEmitted, experimental_assertEmittedOk, type EmitFieldSchema, type EmitObjectSchema, type EmitPropertySchema, type EmitTrackSchema, type EmitToolDefinition, type ExperimentalEmitTool, } from "./experimental-emit.js";
-//# sourceMappingURL=experimental.d.ts.map

package/dist/experimental.js

@@ -1,44 +0,0 @@
-"use strict";
-Object.defineProperty(exports, "__esModule", { value: true });
-exports.experimental_assertEmittedOk = exports.experimental_parseEmitted = exports.experimental_emitTool = exports.experimental_makeDockerRuntime = exports.experimental_dockerRuntime = exports.experimental_withServices = exports.experimental_startServices = void 0;
-/**
- * `vigiles/experimental` — ⚠️ EXPERIMENTAL, UNSTABLE public surface.
- *
- * Everything re-exported here is a DRAFT. It is deliberately quarantined behind
- * the `experimental` subpath (and the `experimental_` name prefix on runtime
- * exports) so the import itself signals the risk at the call site:
- *
- *   import { experimental_startServices } from "vigiles/experimental";
- *
- * NOT covered by the stability guarantee (STABILITY.md): the shape may change or
- * be removed WITHOUT a major-version bump. Do not depend on it in production.
- *
- * Current contents:
- *  - the R3 disposable-service tier (real side-effect testing; see
- *    docs/measuring-skills.md § Experimental and src/services.ts);
- *  - the EMIT delivery for a typed result (`src/experimental-emit.ts`) — a skill
- *    that CALLS a tool with its outcome instead of ending its turn with a fenced
- *    block, which is how an UNFORKED skill can carry an `OutputContract` at all.
- *    Read that module's header before using it: it lists, by number, what is
- *    unproven and what would have to be true to drop the prefix.
- *
- * ⚠️ SAFETY: R3 runs a model-driven skill FOR REAL. The disposable container is
- * the ONLY isolation vigiles provides — it does not confine the skill's filesystem
- * or network. Run it in a disposable environment with NO production access and
- * keep real credentials out of the run. See the SAFETY note in src/services.ts and
- * docs/measuring-skills.md § Experimental.
- *
- * @experimental
- * @module vigiles/experimental
- */
-var services_js_1 = require("./services.js");
-Object.defineProperty(exports, "experimental_startServices", { enumerable: true, get: function () { return services_js_1.experimental_startServices; } });
-Object.defineProperty(exports, "experimental_withServices", { enumerable: true, get: function () { return services_js_1.experimental_withServices; } });
-var services_docker_js_1 = require("./services-docker.js");
-Object.defineProperty(exports, "experimental_dockerRuntime", { enumerable: true, get: function () { return services_docker_js_1.experimental_dockerRuntime; } });
-Object.defineProperty(exports, "experimental_makeDockerRuntime", { enumerable: true, get: function () { return services_docker_js_1.experimental_makeDockerRuntime; } });
-var experimental_emit_js_1 = require("./experimental-emit.js");
-Object.defineProperty(exports, "experimental_emitTool", { enumerable: true, get: function () { return experimental_emit_js_1.experimental_emitTool; } });
-Object.defineProperty(exports, "experimental_parseEmitted", { enumerable: true, get: function () { return experimental_emit_js_1.experimental_parseEmitted; } });
-Object.defineProperty(exports, "experimental_assertEmittedOk", { enumerable: true, get: function () { return experimental_emit_js_1.experimental_assertEmittedOk; } });
-//# sourceMappingURL=experimental.js.map