chore(package): remove deprecated @agentv/eval package

.agents/publish.md

@@ -32,6 +32,7 @@ Contract gates:
 ## Published Packages
 
 - `packages/core/` publishes as `@agentv/core`
+- `packages/sdk/` publishes as `@agentv/sdk`
 - `apps/cli/` publishes as `agentv`
 - The CLI bundles workspace dependencies via tsup with `noExternal: ["@agentv/core"]`
 - Install with `bun install -g agentv` or `npm install -g agentv`

Dockerfile

@@ -5,7 +5,6 @@ WORKDIR /app
 COPY package.json bun.lock ./
 COPY packages/core/package.json packages/core/
 COPY packages/sdk/package.json packages/sdk/
-COPY packages/eval/package.json packages/eval/
 COPY apps/cli/package.json apps/cli/
 COPY apps/dashboard/package.json apps/dashboard/
 COPY apps/web/package.json apps/web/
@@ -60,8 +59,6 @@ COPY --from=build /app/packages/core/dist ./packages/core/dist
 COPY --from=build /app/packages/core/package.json ./packages/core/
 COPY --from=build /app/packages/sdk/dist ./packages/sdk/dist
 COPY --from=build /app/packages/sdk/package.json ./packages/sdk/
-COPY --from=build /app/packages/eval/dist ./packages/eval/dist
-COPY --from=build /app/packages/eval/package.json ./packages/eval/
 COPY --from=build /app/apps/cli/dist ./apps/cli/dist
 COPY --from=build /app/apps/cli/package.json ./apps/cli/
 COPY --from=build /app/apps/cli/node_modules ./apps/cli/node_modules

apps/web/src/content/docs/docs/evaluation/sdk.mdx

@@ -35,7 +35,7 @@ npm install @agentv/sdk
 import { defineCodeGrader } from '@agentv/sdk';
 ```
 
-The general policy is hard convergence for same-week or unreleased surface names: use the correct package, field, or wire name instead of carrying aliases. The package rename is the exception because `@agentv/eval` was already published. It remains a temporary deprecated compatibility package that re-exports `@agentv/sdk` for existing consumers, but it should not appear in new docs, examples, scaffolds, or skills except as migration guidance.
+The general policy is hard convergence for same-week or unreleased surface names: use the correct package, field, or wire name instead of carrying aliases. `@agentv/eval` was already published, then deprecated on npm, and has been removed from this repository. New docs, examples, scaffolds, and skills should use `@agentv/sdk` directly.
 
 ## Choose a Surface
 

docs/adr/2026-06-18-sdk-surface-decision.md

@@ -4,6 +4,10 @@ Date: 2026-06-18
 
 Status: Accepted
 
+Update 2026-06-22: `@agentv/eval` has been deprecated on npm and removed from
+this repository. `@agentv/sdk` is the only lightweight TypeScript SDK package
+published by the release workflow.
+
 Supersedes: the earlier 2026-06-18 decision in this file that rejected a
 separate `@agentv/sdk` package.
 
@@ -61,13 +65,14 @@ For this package rename, npm evidence changes the compatibility choice:
   2026-05-19 through 2026-06-17.
 - `@agentv/sdk` is not yet published at the time of this decision.
 
-Therefore `@agentv/eval` remains only as a thin deprecated compatibility
-package that re-exports `@agentv/sdk` for existing consumers. It should not be
-used by new docs, examples, scaffolds, or skills except when explaining the
-migration.
+Therefore `@agentv/eval` was kept temporarily as a thin deprecated
+compatibility package that re-exported `@agentv/sdk` for existing consumers. It
+must not be used by new docs, examples, scaffolds, or skills except when
+explaining the migration.
 
-Future removal of `@agentv/eval` requires an explicit release/migration
-decision. The compatibility package must not grow new API surface.
+After npm deprecation, the explicit removal decision was made on 2026-06-22.
+The compatibility package is no longer part of the workspace, release script,
+publish script, or runtime Docker image.
 
 ## Non-Goals
 
@@ -95,15 +100,12 @@ Positive:
 
 Negative:
 
-- one temporary compatibility package remains until a later removal decision
-- release scripts and examples need to carry both package paths during the
-  migration window
+- users of the deprecated package must migrate imports to `@agentv/sdk`
 
 ## Tracker Impact
 
 - `av-bv4.11`: this ADR supersedes the previous no-new-sdk decision and records
   the new package-boundary decision.
-- `av-bv4.12`: implementation should move the SDK surface to `packages/sdk` /
-  `@agentv/sdk`, keep `@agentv/eval` only as a deprecated shim because it was
-  already published, and update repo references so the old name is not taught as
-  primary.
+- `av-bv4.12`: implementation moved the SDK surface to `packages/sdk` /
+  `@agentv/sdk`. The deprecated shim has since been removed after npm
+  deprecation.

docs/plans/2026-06-06-001-agentv-eval-authoring-extensibility-plan.md

@@ -23,7 +23,7 @@ Prior research on `av-r0s` found that AgentV already has the right low-level com
 - Custom assertions are discovered from `.agentv/assertions/` in `packages/core/src/evaluation/registry/assertion-discovery.ts`.
 - Custom graders are discovered from `.agentv/graders/` in `packages/core/src/evaluation/registry/grader-discovery.ts`.
 - `agentv create` already scaffolds evals, assertions, and providers in `apps/cli/src/commands/create/commands.ts`.
-- The lightweight SDK contract lives in `packages/eval/src/assertion.ts`.
+- The lightweight SDK contract lives in `packages/sdk/src/assertion.ts`.
 
 The main ceremony problem is not that the schema cannot represent tasks. It is that users must repeatedly hand-write the same layout, provenance metadata, adapter scripts, and integration glue.
 

docs/plans/trace-evaluation-architecture.md

@@ -188,7 +188,7 @@ The exact schema belongs in implementation, but these concepts should be stable:
 ### U1. Trace Artifact Model
 
 - **Goal:** Introduce the core TypeScript model, Zod validation, and snake_case boundary conversion for trace artifacts.
-- **Files:** `packages/core/src/evaluation/trace.ts`, `packages/core/src/evaluation/types.ts`, `packages/eval/src/schemas.ts`, new focused files under `packages/core/src/evaluation/trace/` if the existing file becomes too large.
+- **Files:** `packages/core/src/evaluation/trace.ts`, `packages/core/src/evaluation/types.ts`, `packages/sdk/src/schemas.ts`, new focused files under `packages/core/src/evaluation/trace/` if the existing file becomes too large.
 - **Patterns:** Follow the existing `TraceSummary`, `TokenUsage`, and project boundary conversion conventions. Keep internal fields camelCase and persisted fields snake_case.
 - **Test Scenarios:** Add tests that validate round-trip conversion, missing optional content, inferred duration flags, branch metadata, and raw evidence handles.
 - **Verification:** Unit tests should prove summaries can be derived from trace artifacts without changing current summary behavior, and that trace artifacts do not embed a separate summary payload.
@@ -268,7 +268,7 @@ The exact schema belongs in implementation, but these concepts should be stable:
 ### U7. Grader Context Upgrade
 
 - **Goal:** Let built-in and code graders receive trace artifacts in addition to compact summaries and output messages.
-- **Files:** `packages/core/src/evaluation/graders/types.ts`, `packages/core/src/evaluation/graders/tool-trajectory.ts`, `packages/core/src/evaluation/graders/execution-metrics.ts`, `packages/core/src/evaluation/graders/code-grader.ts`, `packages/eval/src/index.ts`, `packages/eval/src/schemas.ts`.
+- **Files:** `packages/core/src/evaluation/graders/types.ts`, `packages/core/src/evaluation/graders/tool-trajectory.ts`, `packages/core/src/evaluation/graders/execution-metrics.ts`, `packages/core/src/evaluation/graders/code-grader.ts`, `packages/sdk/src/index.ts`, `packages/sdk/src/schemas.ts`.
 - **Patterns:** Keep existing graders that only read `trace` or `output` working. Trace-aware graders use the richer object.
 - **Test Scenarios:** Existing `tool-trajectory` modes should pass from live output and from trace artifact input. Argument matching, ordering, latency, status/error matching, and evidence text should be covered.
 - **Verification:** `trace score` should run `tool-trajectory` against imported traces, not only metrics-only graders.

package.json

@@ -6,15 +6,15 @@
   "packageManager": "bun@1.3.3",
   "workspaces": ["apps/*", "packages/*"],
   "scripts": {
-    "build": "bun --filter @agentv/core build && bun --filter @agentv/sdk build && bun --filter @agentv/eval build && bun --filter @agentv/phoenix-adapter build && bun --filter @agentv/dashboard build && bun --filter agentv build",
+    "build": "bun --filter @agentv/core build && bun --filter @agentv/sdk build && bun --filter @agentv/phoenix-adapter build && bun --filter @agentv/dashboard build && bun --filter agentv build",
     "verify": "bun run build && bun run typecheck && bun run lint && bun run test",
-    "typecheck": "bun --filter @agentv/core typecheck && bun --filter @agentv/sdk typecheck && bun --filter @agentv/eval typecheck && bun --filter @agentv/phoenix-adapter typecheck && bun --filter agentv typecheck",
+    "typecheck": "bun --filter @agentv/core typecheck && bun --filter @agentv/sdk typecheck && bun --filter @agentv/phoenix-adapter typecheck && bun --filter agentv typecheck",
     "typecheck:workspace": "tsc -b tsconfig.build.json",
     "typecheck:watch": "bun --filter @agentv/core typecheck -- --watch & bun --filter agentv typecheck -- --watch",
     "lint": "biome check .",
     "format": "biome format --write .",
     "fix": "biome check --write .",
-    "test": "bun --filter @agentv/core test && bun --filter @agentv/sdk test && bun --filter @agentv/eval test && bun --filter @agentv/phoenix-adapter test && bun --filter agentv test && bun --filter @agentv/dashboard test",
+    "test": "bun --filter @agentv/core test && bun --filter @agentv/sdk test && bun --filter @agentv/phoenix-adapter test && bun --filter agentv test && bun --filter @agentv/dashboard test",
     "test:watch": "bun --filter @agentv/core test:watch & bun --filter agentv test:watch",
     "agentv": "bun apps/cli/src/cli.ts",
     "agentv:buildrun": "bun run build && bun apps/cli/dist/cli.js",

packages/sdk/README.md

@@ -21,7 +21,7 @@ npm install @agentv/sdk
 import { defineCodeGrader } from '@agentv/sdk';
 ```
 
-`@agentv/eval` remains only as a temporary deprecated compatibility package that re-exports this SDK for existing consumers. New docs, examples, scaffolds, and skills should not import from it.
+`@agentv/eval` was a temporary deprecated compatibility package for this SDK. It is no longer published from this repository. Use `@agentv/sdk` directly.
 
 ## Quick Start
 

scripts/publish.ts

@@ -25,7 +25,7 @@ if (requestedTag !== undefined && requestedTag !== 'next') {
 const npmTag = requestedTag ?? 'latest';
 const publishArgs = ['--tag', npmTag, '--access', 'public'];
 
-const PACKAGES = ['packages/core', 'packages/sdk', 'packages/eval', 'apps/cli'];
+const PACKAGES = ['packages/core', 'packages/sdk', 'apps/cli'];
 
 interface PackageJson {
   name: string;

scripts/release.ts

@@ -35,7 +35,6 @@ const NEXT_PRERELEASE_TAG = 'next';
 const PACKAGE_PATHS = [
   'packages/core/package.json',
   'packages/sdk/package.json',
-  'packages/eval/package.json',
   'apps/cli/package.json',
 ];
 

skills-data/agentv-eval-writer/SKILL.md

@@ -18,7 +18,7 @@ Comprehensive docs: https://agentv.dev
 
 Treat YAML as the canonical portable model. Prefer authoring `.eval.yaml` / `EVAL.yaml` first, then use TypeScript helpers, Python scripts, or executable graders only when they lower to the same fields or when the evaluation logic must actually run code.
 
-Use `@agentv/sdk` for TypeScript helper imports. Do not use `@agentv/eval` for new evals, examples, scaffolds, or skill guidance; it is only a deprecated compatibility shim for existing consumers during migration.
+Use `@agentv/sdk` for TypeScript helper imports. Do not use `@agentv/eval` for new evals, examples, scaffolds, or skill guidance; it was a deprecated compatibility package and has been removed from this repository.
 
 ## Evaluation Types