agentv 4.44.0-next.1 → 5.1.0-next.1

package/README.md

@@ -1,36 +1,8 @@
 # AgentV
 
-**Evaluate AI agents against real repos from the terminal. No server. No signup.**
+Test AI targets on real repo tasks and measure what actually works.
 
-```bash
-npm install -g agentv
-agentv init
-agentv eval evals/example.yaml
-```
-
-That's it. Results in seconds, not minutes.
-
-## What it does
-
-AgentV runs evaluation cases against your AI agents and scores them with deterministic code graders + customizable LLM graders. Everything lives in Git — YAML eval files, markdown judge prompts, JSONL results.
-
-```yaml
-# evals/math.yaml
-description: Math problem solving
-tests:
-  - id: addition
-    input: What is 15 + 27?
-    expected_output: "42"
-    assertions:
-      - type: contains
-        value: "42"
-```
-
-```bash
-agentv eval evals/math.yaml
-```
-
-## Why AgentV?
+## Why?
 
 - **Local-first** — runs on your machine, no cloud accounts or API keys for eval infrastructure
 - **Repo-backed workspaces** — reuse real repos, setup scripts, and existing harnesses instead of rebuilding synthetic tasks
@@ -38,7 +10,18 @@ agentv eval evals/math.yaml
 - **Version-controlled** — evals, judges, and results all live in Git
 - **Hybrid graders** — deterministic code checks + LLM-based subjective scoring
 - **CI/CD native** — exit codes, JSONL output, threshold flags for pipeline gating
-- **Any agent** — supports Claude, Codex, Copilot, VS Code, Pi, Azure OpenAI, or any CLI agent
+- **Any target** — run against agents, model providers, gateways, replay targets, CLI wrappers, transcript providers, and future app or service wrappers
+
+## Core Concepts
+
+- **Eval suite / imports / tests** are the task corpus: the prompts, cases, datasets, and imported benchmarks you want to evaluate.
+- **Category** is derived from where the eval lives, such as folder path and file name. Use paths to organize the corpus instead of repeating category labels in every eval.
+- **Workspace / fixtures / graders** are task-owned context: repos, setup scripts, files, fixtures, isolation, deterministic checks, and LLM grading prompts.
+- **Target** is the system under test: an agent, provider, gateway, replay target, CLI wrapper, transcript provider, or future app/service wrapper. Each eval selects one `target` by configured target `id` or with an eval-local target object.
+- **Tags** are run/result grouping labels. `tags.experiment` is the default experiment namespace, such as `with-skills` or `without-skills`; keep suite/category and target/model names out of that tag.
+- **Evaluate options** configure eval run behavior such as `max_concurrency`, repeat policy, and budgets.
+- **Default test** configures inherited per-test defaults such as score `threshold`.
+- **Run** is one concrete execution of a tagged eval against a resolved target that writes portable artifacts for readers such as Dashboard, compare, and trend.
 
 ## Quick start
 
@@ -48,54 +31,208 @@ npm install -g agentv
 agentv init
 ```
 
-**2. Configure targets** in `.agentv/targets.yaml` — point to your agent or LLM provider.
+**2. Configure targets and graders** in `.agentv/config.yaml` — point to the system under test and the reusable grader. Provider settings live under `config`, and target `id` is the selection name used by evals and CLI flags:
+
+```yaml
+targets:
+  - id: local-openai
+    provider: openai
+    runtime: host
+    config:
+      api_format: chat
+      base_url: ${{ LOCAL_OPENAI_PROXY_BASE_URL }}
+      api_key: ${{ LOCAL_OPENAI_PROXY_API_KEY }}
+      model: ${{ LOCAL_OPENAI_PROXY_MODEL }}
+
+graders:
+  - id: local-openai-grader
+    provider: openai
+    config:
+      api_format: chat
+      base_url: ${{ LOCAL_OPENAI_PROXY_BASE_URL }}
+      api_key: ${{ LOCAL_OPENAI_PROXY_API_KEY }}
+      model: ${{ LOCAL_OPENAI_PROXY_MODEL }}
+
+defaults:
+  target: local-openai
+  grader: local-openai-grader
+```
+
+**3. Create shared test defaults** in `evals/default-test.yaml`. This is a partial test config that AgentV applies to each test:
+
+```yaml
+threshold: 0.8
+options:
+  rubric_prompt: |
+    You are an expert grader. Evaluate the candidate answer against each rubric item.
+    Award credit only when the answer directly supports the criterion.
+
+    [[ ## question ## ]]
+    {{ input }}
+
+    [[ ## rubric ## ]]
+    {{ rubrics }}
+
+    [[ ## answer ## ]]
+    {{ output }}
+```
 
-**3. Create an eval** in `evals/`:
+**4. Create an eval** in `evals/my-eval.eval.yaml`:
 ```yaml
 description: Code generation quality
+tags:
+  experiment: with-skills
+target: local-openai
+evaluate_options:
+  max_concurrency: 2
+
+default_test: file://./default-test.yaml
+
+prompts:
+  - "{{ input }}"
+
 tests:
   - id: fizzbuzz
-    criteria: Write a correct FizzBuzz implementation
-    input: Write FizzBuzz in Python
-    assertions:
+    vars:
+      input: Write FizzBuzz in Python. Use lowercase output strings "fizz", "buzz", and "fizzbuzz". Return only one Python code block.
+    assert:
       - type: contains
         value: "fizz"
-      - type: code-grader
-        command: ./validators/check_syntax.py
-      - type: llm-grader
-        prompt: ./graders/correctness.md
+      - Implements correct FizzBuzz logic for multiples of 3, 5, and 15
+      - type: script
+        command: ["python3", "../validators/check_syntax.py"]
+      - type: llm-rubric
+        value:
+          - outcome: Solution is simple and idiomatic Python
+            weight: 0.5
+          - outcome: Handles the 3, 5, and 15 branches correctly
+            weight: 1.5
+```
+
+Plain assertion strings are short-form rubric criteria: AgentV groups them into
+`llm-rubric` and writes each criterion to `grading.json.assertion_results` for the
+Dashboard. Use explicit `type: llm-rubric` when you need weights, required flags, or
+`score_ranges`, or when you need a custom grader prompt, grader target, or
+output transforms; use string `value` for free-form rubric checks. Executable
+graders use `type: script`.
+
+The target can be an eval-local object when this eval needs target settings of its own:
+
+```yaml
+description: Code generation quality with eval-local target settings
+tags:
+  experiment: with-skills
+target:
+  id: local-mini
+  provider: openai
+  runtime: host
+  config:
+    api_format: chat
+    base_url: ${{ LOCAL_OPENAI_PROXY_BASE_URL }}
+    api_key: ${{ LOCAL_OPENAI_PROXY_API_KEY }}
+    model: gpt-5.4-mini
+evaluate_options:
+  repeat:
+    count: 2
+    strategy: pass_any
+
+default_test:
+  threshold: 0.85
+
+prompts:
+  - "{{ input }}"
+
+tests:
+  - id: fizzbuzz
+    vars:
+      input: Write FizzBuzz in Python
+```
+
+`target: local-openai` resolves the configured target id from `.agentv/config.yaml` and uses its provider, model, hooks, and provider settings. The object form above defines a full eval-local target and must include enough provider configuration to run. AgentV records the resolved target information in run artifacts so results can be audited and replayed. The `tags.experiment` label stays `with-skills` because the condition is unchanged; the model/provider variation belongs to the resolved target metadata.
+
+Use `default_test.threshold` for the inherited per-test pass cutoff. `default_test` can also point at a shared file:
+
+```yaml
+default_test: file://{{ env.AGENTV_REPO_ROOT }}/.agentv/default-test.yaml
+```
+
+AgentV makes `AGENTV_REPO_ROOT` available during eval/config interpolation. Projects that prefer a short name can define their own reference in `.agentv/config.yaml`; `global-default` below is just an example key:
+
+```yaml
+refs:
+  global-default: file://{{ env.AGENTV_REPO_ROOT }}/.agentv/default-test.yaml
 ```
 
-**4. Run it:**
+Then eval files in that project can use `default_test: ref://global-default`.
+
+The checked-in version of this quickstart lives in [`examples/features/readme-quickstart/`](examples/features/readme-quickstart/).
+
+**5. Run it:**
 ```bash
-agentv eval evals/my-eval.yaml
+agentv eval evals/my-eval.eval.yaml
 ```
 
-**5. Compare results across targets:**
+**6. Compare two runs** (pass two run indexes — e.g. before and after a change):
 ```bash
-agentv compare .agentv/results/default/<timestamp>/index.jsonl
+agentv results compare .agentv/results/<baseline-run-id>/.internal/index.jsonl .agentv/results/<candidate-run-id>/.internal/index.jsonl
 ```
 
-## Output formats
+## Results
+
+Each run writes a portable bundle directly under `.agentv/results/<run_id>/`. In this example, `tags.experiment: with-skills` names the condition being measured and `target: local-openai` selects the system under test from `.agentv/config.yaml`; both are recorded as metadata, not path segments. The `.internal/index.jsonl` file is the portable row index used by scripts, CI, and `agentv results compare`; per-case sidecars include the resolved eval and target configuration used for the run.
 
 ```bash
-agentv eval evals/my-eval.yaml --output ./run   # writes ./run/index.jsonl
-cat ./run/index.jsonl                         # JSONL results for scripts/CI
+agentv eval evals/my-eval.eval.yaml
+cat .agentv/results/<run_id>/.internal/index.jsonl
+```
+
+Run bundle layout:
+
+```
+.agentv/results/
+├── 2026-06-30T08-30-00-000Z/     # <run_id> — one committed run bundle
+│   ├── summary.json              # run rollup: metadata, pass rate, counts, cost
+│   ├── fizzbuzz--a1b2c3d4/       # <result_dir> for one test/target row
+│   │   ├── summary.json          # optional per-case rollup across attempts
+│   │   ├── test/                 # generated test bundle: frozen inputs for reproducibility
+│   │   │   ├── EVAL.yaml         #   resolved eval spec
+│   │   │   ├── targets.yaml      #   resolved target config
+│   │   │   └── graders/          #   grader files used
+│   │   └── sample-1/             # one materialized sample
+│   │       ├── result.json       # compact attempt manifest
+│   │       ├── grading.json      # assertion_results and grader evidence
+│   │       ├── metrics.json      # tool calls, transcript stats, behavior metrics
+│   │       ├── transcript.json        # normalized agent transcript
+│   │       ├── transcript-raw.jsonl   # raw agent output (debugging)
+│   │       └── outputs/          # captured stdout and grader outputs
+│   └── .internal/
+│       └── index.jsonl           # row index for scripts/CI and `agentv results compare`
+├── .indexes/                     # reserved local/rebuildable indexes
+└── .cache/                       # reserved local cache
 ```
 
 ## TypeScript SDK
 
-Use AgentV programmatically:
+Use `evaluate()` when your application owns the run:
 
 ```typescript
 import { evaluate } from '@agentv/sdk';
 
 const { results, summary } = await evaluate({
+  experiment: 'with-skills',
+  task: async (input) => runMyAppTarget(input),
+  threshold: 0.8,
+  prompts: ['{{ input }}'],
   tests: [
     {
-      id: 'greeting',
-      input: 'Say hello',
-      assertions: [{ type: 'contains', value: 'Hello' }],
+      id: 'fizzbuzz',
+      vars: { input: 'Write FizzBuzz in Python' },
+      assert: [
+        { type: 'contains', value: 'fizz' },
+        'Implements correct FizzBuzz logic for multiples of 3, 5, and 15',
+        { type: 'script', command: ['python3', './validators/check_syntax.py'] },
+        { type: 'llm-rubric', value: ['Solution is simple and idiomatic Python'] },
+      ],
     },
   ],
 });
@@ -103,12 +240,57 @@ const { results, summary } = await evaluate({
 console.log(`${summary.passed}/${summary.total} passed`);
 ```
 
+Use `defineEval()` when you want AgentV to run the TypeScript eval file:
+
+```typescript
+import { defineEval } from '@agentv/sdk';
+
+export default defineEval({
+  description: 'Code generation quality',
+  tags: { experiment: 'with-skills' },
+  target: {
+    extends: 'copilot-sdk',
+    model: 'claude-sonnet-4.6',
+  },
+  repeat: {
+    count: 3,
+    strategy: 'pass_any',
+    earlyExit: false,
+  },
+  threshold: 0.8,
+  prompts: ['{{ input }}'],
+  workspace: {
+    scope: 'attempt',
+    repos: [
+      {
+        path: './fixture',
+        repo: 'EntityProcess/agentv-contract-fixture',
+        commit: '21a34daed7ebcfe36cbed053607622a55e5e94cb',
+      },
+    ],
+  },
+  tests: [
+    {
+      id: 'fizzbuzz',
+      vars: { input: 'Write FizzBuzz in Python' },
+      assert: [
+        { type: 'contains', value: 'fizz' },
+        'Implements correct FizzBuzz logic for multiples of 3, 5, and 15',
+        { type: 'script', command: ['python3', './validators/check_syntax.py'] },
+        { type: 'llm-rubric', value: ['Solution is simple and idiomatic Python'] },
+      ],
+    },
+  ],
+});
+```
+
 ## Documentation
 
 Full docs at [agentv.dev/docs](https://agentv.dev/docs/getting-started/introduction/).
 
 - [Eval files](https://agentv.dev/docs/evaluation/eval-files/) — format and structure
-- [Custom graders](https://agentv.dev/docs/graders/custom-graders/) — code graders in any language
+- [Custom assertions](https://agentv.dev/docs/graders/custom-assertions/) — reusable assertion types
+- [Script graders](https://agentv.dev/docs/graders/script-graders/) — command-backed graders in any language
 - [Rubrics](https://agentv.dev/docs/evaluation/rubrics/) — structured criteria scoring
 - [Targets](https://agentv.dev/docs/targets/configuration/) — configure agents and providers
 - [Compare results](https://agentv.dev/docs/tools/compare/) — A/B testing and regression detection
@@ -136,8 +318,8 @@ AGENTV_RESULTS_REPO=EntityProcess/agentv-evalresults \
 ```
 
 The script clones AgentV examples into `~/agentv-dashboard`, clones the results
-repo, writes the Dashboard project registry under `$AGENTV_HOME/config.yaml`,
-builds the Docker image, and starts Dashboard at `http://localhost:3117`.
+repo, writes the Dashboard project registry under the `$AGENTV_HOME` config
+pair, builds the Docker image, and starts Dashboard at `http://localhost:3117`.
 
 ## License
 

package/dist/{agentv-provider-AYXH7WLW-NJRC6UQX.js → agentv-provider-AYXH7WLW-6YZKZ7PS.js}

@@ -2,8 +2,8 @@ import { createRequire } from 'node:module'; const require = createRequire(impor
 import {
   AgentvProvider
 } from "./chunk-M7BUKBAF.js";
-import "./chunk-5H446C7X.js";
+import "./chunk-QMRVH5ZP.js";
 export {
   AgentvProvider
 };
-//# sourceMappingURL=agentv-provider-AYXH7WLW-NJRC6UQX.js.map
+//# sourceMappingURL=agentv-provider-AYXH7WLW-6YZKZ7PS.js.map

package/dist/{artifact-writer-72MQQELS.js → artifact-writer-IORG7CW5.js}

@@ -4,41 +4,50 @@ import {
   buildResultIndexArtifact,
   writeArtifactsFromResults,
   writePerTestArtifacts
-} from "./chunk-KAGZRUO6.js";
-import "./chunk-NCNDA3F6.js";
-import "./chunk-76FOHROU.js";
-import "./chunk-BPGJ4HBU.js";
+} from "./chunk-CBBBRTQ3.js";
+import "./chunk-GJWPQXHN.js";
+import "./chunk-PEUTJS7B.js";
 import {
   RESULT_INDEX_FILENAME,
+  RUN_CONFIG_FILENAME,
+  RUN_SUMMARY_FILENAME,
   aggregateRunDir,
   buildAggregateGradingArtifact,
-  buildBenchmarkArtifact,
+  buildEvalTestTargetKey,
+  buildEvaluationResultTargetKey,
   buildGradingArtifact,
+  buildRunSummaryArtifact,
   buildTestTargetKey,
   buildTimingArtifact,
   deduplicateByTestIdTarget,
   parseJsonlResults,
+  readRunConfigArtifact,
   writeArtifacts,
-  writeInitialBenchmarkArtifact
-} from "./chunk-5DYGN547.js";
-import "./chunk-NPVGBFF6.js";
+  writeInitialRunSummaryArtifact
+} from "./chunk-ZGPCX334.js";
+import "./chunk-7BGERE6L.js";
 import "./chunk-M7BUKBAF.js";
-import "./chunk-5H446C7X.js";
+import "./chunk-QMRVH5ZP.js";
 export {
   RESULT_INDEX_FILENAME,
+  RUN_CONFIG_FILENAME,
+  RUN_SUMMARY_FILENAME,
   aggregateRunDir,
   buildAggregateGradingArtifact,
-  buildBenchmarkArtifact,
+  buildEvalTestTargetKey,
+  buildEvaluationResultTargetKey,
   buildGradingArtifact,
   buildIndexArtifactEntry,
   buildResultIndexArtifact,
+  buildRunSummaryArtifact,
   buildTestTargetKey,
   buildTimingArtifact,
   deduplicateByTestIdTarget,
   parseJsonlResults,
+  readRunConfigArtifact,
   writeArtifacts,
   writeArtifactsFromResults,
-  writeInitialBenchmarkArtifact,
+  writeInitialRunSummaryArtifact,
   writePerTestArtifacts
 };
-//# sourceMappingURL=artifact-writer-72MQQELS.js.map
+//# sourceMappingURL=artifact-writer-IORG7CW5.js.map

package/dist/{chunk-NPVGBFF6.js → chunk-7BGERE6L.js}

@@ -1,6 +1,6 @@
 import { createRequire } from 'node:module'; const require = createRequire(import.meta.url);
 
-// ../../packages/core/dist/chunk-3WGHC7LC.js
+// ../../packages/core/dist/chunk-F4TEFL4T.js
 function shellEscapePath(value) {
   if (process.platform === "win32") {
     return `"${value.replaceAll('"', '""')}"`;
@@ -11,49 +11,8 @@ async function execFileWithStdin(argv, stdinPayload, options = {}) {
   if (argv.length === 0) {
     throw new Error("Executable argv must include at least one entry");
   }
-  if (typeof Bun !== "undefined") {
-    return execFileWithStdinBun(argv, stdinPayload, options);
-  }
   return execFileWithStdinNode(argv, stdinPayload, options);
 }
-async function execFileWithStdinBun(argv, stdinPayload, options) {
-  const command = [...argv];
-  const encoder = new TextEncoder();
-  const proc = Bun.spawn(command, {
-    cwd: options.cwd,
-    stdin: encoder.encode(stdinPayload),
-    stdout: "pipe",
-    stderr: "pipe",
-    // Merge additional env vars with process.env
-    env: options.env ? { ...process.env, ...options.env } : process.env
-  });
-  let timedOut = false;
-  const timeout = options.timeoutMs !== void 0 ? setTimeout(() => {
-    timedOut = true;
-    proc.kill("SIGKILL");
-  }, options.timeoutMs) : void 0;
-  try {
-    const stdoutPromise = proc.stdout ? new Response(proc.stdout).text() : Promise.resolve("");
-    const stderrPromise = proc.stderr ? new Response(proc.stderr).text() : Promise.resolve("");
-    const [stdout, stderr, exitCode] = await Promise.all([
-      stdoutPromise,
-      stderrPromise,
-      proc.exited
-    ]);
-    if (timedOut) {
-      throw new Error(`Process timed out after ${options.timeoutMs}ms`);
-    }
-    return {
-      stdout: stdout.replace(/\r\n/g, "\n"),
-      stderr: stderr.replace(/\r\n/g, "\n"),
-      exitCode
-    };
-  } finally {
-    if (timeout !== void 0) {
-      clearTimeout(timeout);
-    }
-  }
-}
 async function execFileWithStdinNode(argv, stdinPayload, options) {
   const { spawn } = await import("node:child_process");
   return new Promise((resolve, reject) => {
@@ -148,4 +107,4 @@ export {
   execFileWithStdin,
   execShellWithStdin
 };
-//# sourceMappingURL=chunk-NPVGBFF6.js.map
+//# sourceMappingURL=chunk-7BGERE6L.js.map

package/dist/chunk-7BGERE6L.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../packages/core/src/runtime/exec.ts"],"sourcesContent":["interface ExecOptions {\n  readonly cwd?: string;\n  readonly timeoutMs?: number;\n  /** Additional environment variables to pass to the subprocess */\n  readonly env?: Record<string, string>;\n}\n\nfunction shellEscapePath(value: string): string {\n  if (process.platform === 'win32') {\n    // Very small escape helper for file paths in cmd.exe context.\n    // Wrap in double-quotes and escape existing double-quotes.\n    return `\"${value.replaceAll('\"', '\"\"')}\"`;\n  }\n  // POSIX: single-quote escape (close/open around embedded single quotes).\n  return `'${value.replaceAll(\"'\", `'\\\"'\\\"'`)}'`;\n}\n\n// IPC protocol: stdin/stdout JSON was chosen over JSON-RPC (overkill for one-shot eval),\n// HTTP (unnecessary network overhead for local scripts), and in-process (no isolation,\n// single-language only). Each evaluation is isolated by design for reproducibility and\n// safe parallelization. Process spawn overhead (~10-50ms) is acceptable for eval workloads.\nexport async function execFileWithStdin(\n  argv: readonly string[],\n  stdinPayload: string,\n  options: ExecOptions = {},\n): Promise<{\n  readonly stdout: string;\n  readonly stderr: string;\n  readonly exitCode: number;\n}> {\n  if (argv.length === 0) {\n    throw new Error('Executable argv must include at least one entry');\n  }\n\n  // Use Node's child_process path even under Bun. Bun 1.3.3 can hang when a\n  // Bun.spawn child is a Node process reading stdin, which breaks script grader\n  // and workspace hook execution on CI's pinned Bun version.\n  return execFileWithStdinNode(argv, stdinPayload, options);\n}\n\n/**\n * Node.js implementation using child_process.spawn\n */\nasync function execFileWithStdinNode(\n  argv: readonly string[],\n  stdinPayload: string,\n  options: ExecOptions,\n): Promise<{\n  readonly stdout: string;\n  readonly stderr: string;\n  readonly exitCode: number;\n}> {\n  const { spawn } = await import('node:child_process');\n\n  return new Promise((resolve, reject) => {\n    const [cmd, ...args] = argv;\n    const child = spawn(cmd, args, {\n      cwd: options.cwd,\n      stdio: ['pipe', 'pipe', 'pipe'],\n      // Merge additional env vars with process.env\n      env: options.env ? { ...process.env, ...options.env } : process.env,\n    });\n\n    const stdoutChunks: Buffer[] = [];\n    const stderrChunks: Buffer[] = [];\n\n    child.stdout?.on('data', (chunk: Buffer) => stdoutChunks.push(chunk));\n    child.stderr?.on('data', (chunk: Buffer) => stderrChunks.push(chunk));\n\n    let timedOut = false;\n    const timeout =\n      options.timeoutMs !== undefined\n        ? setTimeout(() => {\n            timedOut = true;\n            child.kill('SIGKILL');\n          }, options.timeoutMs)\n        : undefined;\n\n    child.on('error', (error) => {\n      if (timeout !== undefined) clearTimeout(timeout);\n      reject(error);\n    });\n\n    child.on('close', (code) => {\n      if (timeout !== undefined) clearTimeout(timeout);\n\n      if (timedOut) {\n        reject(new Error(`Process timed out after ${options.timeoutMs}ms`));\n        return;\n      }\n\n      const stdout = Buffer.concat(stdoutChunks).toString('utf8').replace(/\\r\\n/g, '\\n');\n      const stderr = Buffer.concat(stderrChunks).toString('utf8').replace(/\\r\\n/g, '\\n');\n\n      resolve({\n        stdout,\n        stderr,\n        exitCode: code ?? 0,\n      });\n    });\n\n    // Write stdin and close\n    if (child.stdin) {\n      child.stdin.write(stdinPayload);\n      child.stdin.end();\n    }\n  });\n}\n\n/**\n * Execute a shell command with the given stdin payload.\n *\n * Why this exists:\n * - Some providers/scripts (notably Node.js) must receive stdin reliably.\n * - In some Bun environments, `Bun.spawn` does not forward stdin to Node correctly.\n * - Capture stdout/stderr via temp files to avoid pipe incompatibilities.\n */\nexport async function execShellWithStdin(\n  command: string,\n  stdinPayload: string,\n  options: ExecOptions = {},\n): Promise<{\n  readonly stdout: string;\n  readonly stderr: string;\n  readonly exitCode: number;\n}> {\n  const { mkdir, readFile, rm, writeFile } = await import('node:fs/promises');\n  const { tmpdir } = await import('node:os');\n  const path = await import('node:path');\n  const { randomUUID } = await import('node:crypto');\n\n  const dir = path.join(tmpdir(), `agentv-exec-${randomUUID()}`);\n  await mkdir(dir, { recursive: true });\n\n  const stdinPath = path.join(dir, 'stdin.txt');\n  const stdoutPath = path.join(dir, 'stdout.txt');\n  const stderrPath = path.join(dir, 'stderr.txt');\n\n  await writeFile(stdinPath, stdinPayload, 'utf8');\n\n  const wrappedCommand =\n    process.platform === 'win32'\n      ? `(${command}) < ${shellEscapePath(stdinPath)} > ${shellEscapePath(stdoutPath)} 2> ${shellEscapePath(stderrPath)}`\n      : `(${command}) < ${shellEscapePath(stdinPath)} > ${shellEscapePath(stdoutPath)} 2> ${shellEscapePath(stderrPath)}`;\n\n  const { spawn } = await import('node:child_process');\n  try {\n    const exitCode = await new Promise<number>((resolve, reject) => {\n      const child = spawn(wrappedCommand, {\n        shell: true,\n        cwd: options.cwd,\n        stdio: ['ignore', 'ignore', 'ignore'],\n        // Merge additional env vars with process.env\n        env: options.env ? { ...process.env, ...options.env } : process.env,\n      });\n\n      const timeout = options.timeoutMs\n        ? setTimeout(() => {\n            child.kill();\n            reject(new Error(`Process timed out after ${options.timeoutMs}ms`));\n          }, options.timeoutMs)\n        : undefined;\n\n      child.on('error', (error) => {\n        if (timeout !== undefined) {\n          clearTimeout(timeout);\n        }\n        reject(error);\n      });\n\n      child.on('exit', (code) => {\n        if (timeout !== undefined) {\n          clearTimeout(timeout);\n        }\n        resolve(code ?? 0);\n      });\n    });\n\n    const stdout = (await readFile(stdoutPath, 'utf8')).replace(/\\r\\n/g, '\\n');\n    const stderr = (await readFile(stderrPath, 'utf8')).replace(/\\r\\n/g, '\\n');\n    return { stdout, stderr, exitCode };\n  } finally {\n    await rm(dir, { recursive: true, force: true });\n  }\n}\n"],"mappings":";;;AAOA,SAAS,gBAAgB,OAAuB;AAC9C,MAAI,QAAQ,aAAa,SAAS;AAGhC,WAAO,IAAI,MAAM,WAAW,KAAK,IAAI,CAAC;EACxC;AAEA,SAAO,IAAI,MAAM,WAAW,KAAK,OAAS,CAAC;AAC7C;AAMA,eAAsB,kBACpB,MACA,cACA,UAAuB,CAAC,GAKvB;AACD,MAAI,KAAK,WAAW,GAAG;AACrB,UAAM,IAAI,MAAM,iDAAiD;EACnE;AAKA,SAAO,sBAAsB,MAAM,cAAc,OAAO;AAC1D;AAKA,eAAe,sBACb,MACA,cACA,SAKC;AACD,QAAM,EAAE,MAAM,IAAI,MAAM,OAAO,oBAAoB;AAEnD,SAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,UAAM,CAAC,KAAK,GAAG,IAAI,IAAI;AACvB,UAAM,QAAQ,MAAM,KAAK,MAAM;MAC7B,KAAK,QAAQ;MACb,OAAO,CAAC,QAAQ,QAAQ,MAAM;;MAE9B,KAAK,QAAQ,MAAM,EAAE,GAAG,QAAQ,KAAK,GAAG,QAAQ,IAAI,IAAI,QAAQ;IAClE,CAAC;AAED,UAAM,eAAyB,CAAC;AAChC,UAAM,eAAyB,CAAC;AAEhC,UAAM,QAAQ,GAAG,QAAQ,CAAC,UAAkB,aAAa,KAAK,KAAK,CAAC;AACpE,UAAM,QAAQ,GAAG,QAAQ,CAAC,UAAkB,aAAa,KAAK,KAAK,CAAC;AAEpE,QAAI,WAAW;AACf,UAAM,UACJ,QAAQ,cAAc,SAClB,WAAW,MAAM;AACf,iBAAW;AACX,YAAM,KAAK,SAAS;IACtB,GAAG,QAAQ,SAAS,IACpB;AAEN,UAAM,GAAG,SAAS,CAAC,UAAU;AAC3B,UAAI,YAAY,OAAW,cAAa,OAAO;AAC/C,aAAO,KAAK;IACd,CAAC;AAED,UAAM,GAAG,SAAS,CAAC,SAAS;AAC1B,UAAI,YAAY,OAAW,cAAa,OAAO;AAE/C,UAAI,UAAU;AACZ,eAAO,IAAI,MAAM,2BAA2B,QAAQ,SAAS,IAAI,CAAC;AAClE;MACF;AAEA,YAAM,SAAS,OAAO,OAAO,YAAY,EAAE,SAAS,MAAM,EAAE,QAAQ,SAAS,IAAI;AACjF,YAAM,SAAS,OAAO,OAAO,YAAY,EAAE,SAAS,MAAM,EAAE,QAAQ,SAAS,IAAI;AAEjF,cAAQ;QACN;QACA;QACA,UAAU,QAAQ;MACpB,CAAC;IACH,CAAC;AAGD,QAAI,MAAM,OAAO;AACf,YAAM,MAAM,MAAM,YAAY;AAC9B,YAAM,MAAM,IAAI;IAClB;EACF,CAAC;AACH;AAUA,eAAsB,mBACpB,SACA,cACA,UAAuB,CAAC,GAKvB;AACD,QAAM,EAAE,OAAO,UAAU,IAAI,UAAU,IAAI,MAAM,OAAO,kBAAkB;AAC1E,QAAM,EAAE,OAAO,IAAI,MAAM,OAAO,SAAS;AACzC,QAAM,OAAO,MAAM,OAAO,WAAW;AACrC,QAAM,EAAE,WAAW,IAAI,MAAM,OAAO,aAAa;AAEjD,QAAM,MAAM,KAAK,KAAK,OAAO,GAAG,eAAe,WAAW,CAAC,EAAE;AAC7D,QAAM,MAAM,KAAK,EAAE,WAAW,KAAK,CAAC;AAEpC,QAAM,YAAY,KAAK,KAAK,KAAK,WAAW;AAC5C,QAAM,aAAa,KAAK,KAAK,KAAK,YAAY;AAC9C,QAAM,aAAa,KAAK,KAAK,KAAK,YAAY;AAE9C,QAAM,UAAU,WAAW,cAAc,MAAM;AAE/C,QAAM,iBACJ,QAAQ,aAAa,UACjB,IAAI,OAAO,OAAO,gBAAgB,SAAS,CAAC,MAAM,gBAAgB,UAAU,CAAC,OAAO,gBAAgB,UAAU,CAAC,KAC/G,IAAI,OAAO,OAAO,gBAAgB,SAAS,CAAC,MAAM,gBAAgB,UAAU,CAAC,OAAO,gBAAgB,UAAU,CAAC;AAErH,QAAM,EAAE,MAAM,IAAI,MAAM,OAAO,oBAAoB;AACnD,MAAI;AACF,UAAM,WAAW,MAAM,IAAI,QAAgB,CAAC,SAAS,WAAW;AAC9D,YAAM,QAAQ,MAAM,gBAAgB;QAClC,OAAO;QACP,KAAK,QAAQ;QACb,OAAO,CAAC,UAAU,UAAU,QAAQ;;QAEpC,KAAK,QAAQ,MAAM,EAAE,GAAG,QAAQ,KAAK,GAAG,QAAQ,IAAI,IAAI,QAAQ;MAClE,CAAC;AAED,YAAM,UAAU,QAAQ,YACpB,WAAW,MAAM;AACf,cAAM,KAAK;AACX,eAAO,IAAI,MAAM,2BAA2B,QAAQ,SAAS,IAAI,CAAC;MACpE,GAAG,QAAQ,SAAS,IACpB;AAEJ,YAAM,GAAG,SAAS,CAAC,UAAU;AAC3B,YAAI,YAAY,QAAW;AACzB,uBAAa,OAAO;QACtB;AACA,eAAO,KAAK;MACd,CAAC;AAED,YAAM,GAAG,QAAQ,CAAC,SAAS;AACzB,YAAI,YAAY,QAAW;AACzB,uBAAa,OAAO;QACtB;AACA,gBAAQ,QAAQ,CAAC;MACnB,CAAC;IACH,CAAC;AAED,UAAM,UAAU,MAAM,SAAS,YAAY,MAAM,GAAG,QAAQ,SAAS,IAAI;AACzE,UAAM,UAAU,MAAM,SAAS,YAAY,MAAM,GAAG,QAAQ,SAAS,IAAI;AACzE,WAAO,EAAE,QAAQ,QAAQ,SAAS;EACpC,UAAA;AACE,UAAM,GAAG,KAAK,EAAE,WAAW,MAAM,OAAO,KAAK,CAAC;EAChD;AACF;","names":[]}