@tangle-network/agent-runtime 0.94.0 → 0.94.1

package/dist/chunk-PJNGPWTA.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/mcp/local-harness.ts"],"sourcesContent":["/**\n *\n * Subprocess wrappers for the local coding-harness CLIs installed in the\n * sandbox image (claude-code, codex, opencode). Used by the in-process\n * delegation executor (`createInProcessExecutor`) so a delegated coding task\n * spawns a real harness on a real git worktree instead of provisioning a\n * sibling sandbox.\n *\n * All harness invocations:\n *   - run with `cwd` set to the worktree\n *   - inherit env from the parent (the MCP server inside the sandbox has\n *     the harness's auth already)\n *   - capture stdout/stderr\n *   - support cancellation via AbortSignal\n *   - enforce a wall-clock timeout\n *\n * @experimental\n */\n\nimport { type ChildProcess, spawn } from 'node:child_process'\nimport { createHash } from 'node:crypto'\nimport { constants } from 'node:fs'\nimport { access, chmod, mkdtemp, rm, symlink, writeFile } from 'node:fs/promises'\nimport { homedir, tmpdir } from 'node:os'\nimport { basename, join } from 'node:path'\nimport type { AgentProfile } from '@tangle-network/agent-interface'\n\n/** Local coding harness available inside the sandbox. */\nexport type LocalHarness = 'claude' | 'codex' | 'opencode'\n\ntype ReasoningEffort = NonNullable<NonNullable<AgentProfile['model']>['reasoningEffort']>\n\nconst codexReasoningEffort: Record<ReasoningEffort, string> = {\n  none: 'none',\n  minimal: 'minimal',\n  low: 'low',\n  medium: 'medium',\n  high: 'high',\n  xhigh: 'xhigh',\n  ultracode: 'xhigh',\n}\n\nfunction codexReasoningArgs(reasoningEffort: ReasoningEffort): string[] {\n  const mapped = codexReasoningEffort[reasoningEffort]\n  if (mapped === undefined) {\n    throw new Error(\n      `harnessInvocation: unsupported Codex reasoning effort ${String(reasoningEffort)}`,\n    )\n  }\n  return ['-c', `model_reasoning_effort=\"${mapped}\"`]\n}\n\n/**\n * Default per-harness command + arg shape. `buildArgs` takes ONLY the task prompt and\n * emits the prompt-only invocation (no model, no system prompt) — the safe default shape\n * the in-process executor's `streamPrompt` drives. `modelArgs` maps a resolved model to\n * the harness's selector flag (every supported harness takes `-m <model>`). The §1.5\n * profile-aware mapper `harnessInvocation` composes these to thread the full\n * supervisor-authored profile (systemPrompt + model) into argv.\n */\nconst HARNESS_INVOCATIONS: Record<\n  LocalHarness,\n  {\n    command: string\n    buildArgs: (taskPrompt: string) => string[]\n    /** Map a resolved model to the harness's model-selector flag. */\n    modelArgs: (model: string) => string[]\n    /** Map portable reasoning effort when the harness exposes a native control. */\n    reasoningArgs?: (reasoningEffort: ReasoningEffort) => string[]\n  }\n> = {\n  claude: {\n    command: 'claude',\n    // `-p` IS headless/print mode; the old `--headless` flag was removed from the CLI.\n    // Permission bypass is an explicit per-run opt-in below, never the public default.\n    buildArgs: (taskPrompt) => ['-p', taskPrompt],\n    modelArgs: (model) => ['-m', model],\n  },\n  codex: {\n    command: 'codex',\n    buildArgs: (taskPrompt) => ['exec', taskPrompt],\n    modelArgs: (model) => ['-m', model],\n    reasoningArgs: codexReasoningArgs,\n  },\n  opencode: {\n    command: 'opencode',\n    buildArgs: (taskPrompt) => ['run', taskPrompt],\n    modelArgs: (model) => ['-m', model],\n  },\n}\n\n/** Result of mapping an `AgentProfile` + task prompt onto a harness invocation. */\nexport interface HarnessInvocation {\n  command: string\n  args: string[]\n}\n\nexport interface HarnessInvocationOptions {\n  /** Allow an unattended Claude process to edit its isolated candidate worktree.\n   *  Ignored by harnesses that do not use Claude's permission prompt. */\n  dangerouslySkipPermissions?: boolean\n  /** Run Codex with benchmark-safe process controls and JSONL usage output.\n   *  Valid only for the Codex harness. */\n  codexReproducible?: boolean\n}\n\nconst CODEX_REPRODUCIBLE_ARGS = [\n  '--ephemeral',\n  '--ignore-rules',\n  '--json',\n  '-c',\n  'approval_policy=\"never\"',\n  '-c',\n  'web_search=\"disabled\"',\n  '-c',\n  'project_doc_max_bytes=0',\n  '-c',\n  'skills.include_instructions=false',\n  '-c',\n  'include_apps_instructions=false',\n  '--disable',\n  'tool_suggest',\n  '-c',\n  'features.multi_agent_v2.root_agent_usage_hint_text=\"\"',\n  '-c',\n  'features.multi_agent_v2.subagent_usage_hint_text=\"\"',\n  '-c',\n  'features.multi_agent_v2.multi_agent_mode_hint_text=\"\"',\n  '--strict-config',\n] as const\n\nfunction buildHarnessArgs(\n  harness: LocalHarness,\n  taskPrompt: string,\n  options: HarnessInvocationOptions = {},\n): string[] {\n  const args = HARNESS_INVOCATIONS[harness].buildArgs(taskPrompt)\n  if (harness === 'claude' && options.dangerouslySkipPermissions) {\n    args.push('--dangerously-skip-permissions')\n  }\n  if (options.codexReproducible) {\n    if (harness !== 'codex') {\n      throw new Error('harnessInvocation: codexReproducible requires the Codex harness')\n    }\n    args.push(...CODEX_REPRODUCIBLE_ARGS)\n  }\n  return args\n}\n\n/**\n * Map a supervisor-authored `AgentProfile` + the per-task prompt onto a concrete harness\n * `command` + `args` (the §1.5 fix). UNLIKE the prompt-only `HARNESS_INVOCATIONS.buildArgs`\n * — which drops both the authored model and the system prompt — this threads the FULL\n * profile payload into argv:\n *\n *  - `profile.prompt.systemPrompt` → the PROMPT channel: a portable, harness-agnostic\n *    default that prepends the system prompt above the task prompt (`<system>\\n\\n<task>`),\n *    so the authored standing instructions reach EVERY harness (none of the three CLIs\n *    expose a portable replace-system-prompt flag for a one-shot non-interactive run).\n *  - `profile.prompt.instructions[]` → appended prompt sections in authored order.\n *  - `profile.model.default` → the harness's `-m <model>` selector.\n *\n * The task prompt alone is the floor; an empty/absent profile yields exactly the legacy\n * `buildArgs(taskPrompt)` shape so existing callers are byte-identical.\n */\nexport function harnessInvocation(\n  harness: LocalHarness,\n  profile: AgentProfile,\n  taskPrompt: string,\n  options: HarnessInvocationOptions = {},\n): HarnessInvocation {\n  const invocation = HARNESS_INVOCATIONS[harness]\n  if (!invocation) {\n    throw new Error(`harnessInvocation: unknown harness ${String(harness)}`)\n  }\n\n  if (options.codexReproducible) {\n    const model = profile.model?.default\n    if (typeof model !== 'string' || model.trim().length === 0) {\n      throw new Error('harnessInvocation: codexReproducible requires profile.model.default')\n    }\n    if (profile.model?.reasoningEffort === undefined) {\n      throw new Error('harnessInvocation: codexReproducible requires profile.model.reasoningEffort')\n    }\n  }\n\n  const systemPrompt = profile.prompt?.systemPrompt\n  const instructions = profile.prompt?.instructions\n  if (\n    instructions !== undefined &&\n    (!Array.isArray(instructions) ||\n      instructions.some((instruction) => typeof instruction !== 'string'))\n  ) {\n    throw new Error('harnessInvocation: profile.prompt.instructions must be an array of strings')\n  }\n  const promptSections = [\n    ...(typeof systemPrompt === 'string' && systemPrompt.trim().length > 0 ? [systemPrompt] : []),\n    ...(instructions?.filter((instruction) => instruction.trim().length > 0) ?? []),\n    taskPrompt,\n  ]\n  const composedPrompt = promptSections.join('\\n\\n')\n\n  const args = buildHarnessArgs(harness, composedPrompt, options)\n\n  const model = profile.model?.default\n  if (typeof model === 'string' && model.length > 0) {\n    args.push(...invocation.modelArgs(model))\n  }\n\n  const reasoningEffort = profile.model?.reasoningEffort\n  if (reasoningEffort !== undefined && invocation.reasoningArgs) {\n    args.push(...invocation.reasoningArgs(reasoningEffort))\n  }\n\n  return { command: invocation.command, args }\n}\n\n/** @experimental */\nexport interface RunLocalHarnessOptions {\n  harness: LocalHarness\n  /** Working directory for the subprocess (typically a worktree path). */\n  cwd: string\n  /** Prompt forwarded as the harness CLI's task argument. */\n  taskPrompt: string\n  /**\n   * Pre-built command + args (e.g. from `harnessInvocation` so the full authored\n   * `AgentProfile` — systemPrompt + model — reaches the harness). When set it OVERRIDES the\n   * default prompt-only `buildArgs(taskPrompt)` path; `command` defaults to the harness's\n   * default binary when only `args` is supplied. When absent the legacy prompt-only shape\n   * is used unchanged.\n   */\n  invocation?: { command?: string; args: ReadonlyArray<string> }\n  /** Allow autonomous Claude edits without an interactive permission prompt.\n   *  Use only when `cwd` is an isolated candidate worktree. */\n  dangerouslySkipPermissions?: boolean\n  /** Isolate Codex from ambient configuration/instructions and require JSONL token usage.\n   *  The invocation should come from `harnessInvocation(..., { codexReproducible: true })`. */\n  codexReproducible?: boolean\n  /** Wall-clock kill deadline (ms). Default 5 min. Subprocess SIGTERMed on expiry. */\n  timeoutMs?: number\n  /** Caller cancellation. SIGTERM is sent on abort. */\n  signal?: AbortSignal\n  /** Override env (defaults to inheriting from the parent). */\n  env?: NodeJS.ProcessEnv\n  /**\n   * Test seam — inject a custom spawner so unit tests can mock the\n   * subprocess without touching the OS. Defaults to node's `child_process.spawn`.\n   */\n  spawn?: (\n    command: string,\n    args: ReadonlyArray<string>,\n    opts: {\n      cwd: string\n      env: NodeJS.ProcessEnv\n      stdio: 'pipe'\n    },\n  ) => ChildProcess\n}\n\n/** Exact aggregate usage emitted by Codex's terminal `turn.completed` JSONL event. */\nexport interface CodexTokenUsage {\n  inputTokens: number\n  cachedInputTokens: number\n  outputTokens: number\n  reasoningOutputTokens: number\n}\n\n/** Isolation settings asserted before a reproducible Codex run is allowed to start. */\nexport interface CodexExecutionPolicy {\n  sessionPersistence: 'ephemeral'\n  userConfig: false\n  rules: false\n  projectInstructions: false\n  skillInstructions: false\n  appInstructions: false\n  toolSuggestions: false\n  multiAgentInstructions: false\n  sandbox: 'workspace-write'\n  permissionProfile: 'agent_runtime_reproducible'\n  approvalPolicy: 'never'\n  shellNetwork: false\n  webSearch: false\n  serviceTier: 'default'\n  shellEnvironment: 'core-filtered'\n  loginShell: false\n  credentialsReadable: false\n  parentRepoRead: false\n  gitMetadata: false\n  temporaryDirectory: 'workspace-private'\n  containerSockets: false\n}\n\n/** Zero-model-call evidence for the exact Codex process about to run. */\nexport interface CodexExecutionEvidence {\n  cliVersion: string\n  effectivePromptSha256: string\n  nonPromptArgsSha256: string\n  controlledConfigSha256: string\n  policy: CodexExecutionPolicy\n}\n\n/** @experimental */\nexport interface LocalHarnessResult {\n  /** OS exit code. `null` when killed before exit. */\n  exitCode: number | null\n  /** Concatenated stdout. */\n  stdout: string\n  /** Concatenated stderr. */\n  stderr: string\n  /** Set when the process exited via signal (timeout / abort). */\n  killedBySignal: NodeJS.Signals | null\n  /** Wall-clock duration ms (spawn → exit). */\n  durationMs: number\n  /** Set when timeoutMs elapsed before exit. */\n  timedOut: boolean\n  /** Present for a reproducible Codex run; parsed from the real terminal JSONL event. */\n  usage?: CodexTokenUsage\n  /** Present for reproducible Codex runs; generated and checked before model execution. */\n  evidence?: CodexExecutionEvidence\n}\n\nconst DEFAULT_TIMEOUT_MS = 5 * 60 * 1000\n\n/**\n * Spawn a local coding harness CLI as a subprocess + collect its output.\n *\n * NOT responsible for parsing the harness's output or extracting a diff —\n * the in-process executor's `streamPrompt` orchestrates `git diff` against\n * the worktree after this resolves. This function is intentionally narrow:\n * spawn, wait, capture, return.\n *\n * Fails loud — throws when:\n *   - `cwd` doesn't exist (subprocess emits ENOENT; surfaced as Error)\n *   - the harness binary is not on PATH (ENOENT)\n *\n * Does NOT throw when:\n *   - the subprocess exits non-zero (`result.exitCode` carries the code)\n *   - the subprocess is aborted / timed out (`result.killedBySignal` /\n *     `result.timedOut` carries the reason)\n *\n * @experimental\n */\nexport async function runLocalHarness(\n  options: RunLocalHarnessOptions,\n): Promise<LocalHarnessResult> {\n  const { harness, cwd, taskPrompt } = options\n  if (options.codexReproducible && harness !== 'codex') {\n    throw new Error('runLocalHarness: codexReproducible requires the Codex harness')\n  }\n  if (options.codexReproducible && process.platform !== 'linux') {\n    throw new Error('runLocalHarness: codexReproducible currently requires Linux')\n  }\n  const timeoutMs = options.timeoutMs ?? DEFAULT_TIMEOUT_MS\n  const spawnImpl = options.spawn ?? spawn\n\n  const invocation = HARNESS_INVOCATIONS[harness]\n  if (!invocation) {\n    throw new Error(`runLocalHarness: unknown harness ${String(harness)}`)\n  }\n\n  const startedAt = Date.now()\n  const command = options.invocation?.command ?? invocation.command\n  const args = options.invocation\n    ? [...options.invocation.args]\n    : buildHarnessArgs(harness, taskPrompt, options)\n  if (options.codexReproducible) assertCodexReproducibleInvocation(command, args)\n\n  const baseEnv = options.env ?? process.env\n  const isolated = options.codexReproducible\n    ? await isolateCodexHome(baseEnv, cwd)\n    : {\n        env: baseEnv,\n        cleanup: async () => undefined,\n        controlledConfigSha256: '',\n        ambientAuth: '',\n        isolatedAuth: '',\n        writeProbe: '',\n      }\n  const env = isolated.env\n\n  try {\n    const evidence = options.codexReproducible\n      ? await collectCodexExecutionEvidence({\n          command,\n          args,\n          cwd,\n          env,\n          spawn: spawnImpl,\n          controlledConfigSha256: isolated.controlledConfigSha256,\n          ambientAuth: isolated.ambientAuth,\n          isolatedAuth: isolated.isolatedAuth,\n          writeProbe: isolated.writeProbe,\n          ...(options.signal ? { signal: options.signal } : {}),\n        })\n      : undefined\n    return await new Promise<LocalHarnessResult>((resolve, reject) => {\n      let child: ChildProcess\n      try {\n        child = spawnImpl(command, args, { cwd, env, stdio: 'pipe' })\n      } catch (err) {\n        reject(err instanceof Error ? err : new Error(String(err)))\n        return\n      }\n\n      // The harness takes its task as an argv arg, not on stdin. Leaving stdin\n      // OPEN makes a non-TTY `opencode run` (and likely the other harnesses)\n      // BLOCK forever waiting on input — zero output, SIGTERM at the wall cap,\n      // empty patch -> \"no candidate passed validation\". Close stdin so the\n      // subprocess sees EOF and proceeds (the `cliExecutor` leaf does the same).\n      child.stdin?.end()\n\n      let stdout = ''\n      let stderr = ''\n      let timedOut = false\n      let settled = false\n\n      const timer =\n        timeoutMs > 0\n          ? setTimeout(() => {\n              timedOut = true\n              if (!child.killed) child.kill('SIGTERM')\n            }, timeoutMs)\n          : null\n      if (timer && typeof (timer as { unref?: () => void }).unref === 'function') {\n        ;(timer as { unref: () => void }).unref()\n      }\n\n      const onAbort = () => {\n        if (!child.killed) child.kill('SIGTERM')\n      }\n      if (options.signal) {\n        if (options.signal.aborted) onAbort()\n        else options.signal.addEventListener('abort', onAbort, { once: true })\n      }\n\n      child.stdout?.on('data', (chunk) => {\n        stdout += String(chunk)\n      })\n      child.stderr?.on('data', (chunk) => {\n        stderr += String(chunk)\n      })\n\n      const finalize = (result: LocalHarnessResult) => {\n        if (settled) return\n        settled = true\n        if (timer) clearTimeout(timer)\n        options.signal?.removeEventListener('abort', onAbort)\n        resolve(result)\n      }\n\n      child.on('error', (err) => {\n        if (settled) return\n        settled = true\n        if (timer) clearTimeout(timer)\n        options.signal?.removeEventListener('abort', onAbort)\n        reject(err)\n      })\n\n      child.on('close', (code, signal) => {\n        try {\n          const usage = options.codexReproducible ? parseCodexTokenUsage(stdout) : undefined\n          finalize({\n            exitCode: code,\n            stdout,\n            stderr: redactCodexHome(stderr, env.CODEX_HOME),\n            killedBySignal: signal,\n            durationMs: Date.now() - startedAt,\n            timedOut,\n            ...(usage ? { usage } : {}),\n            ...(evidence ? { evidence } : {}),\n          })\n        } catch (err) {\n          if (settled) return\n          settled = true\n          if (timer) clearTimeout(timer)\n          options.signal?.removeEventListener('abort', onAbort)\n          reject(err instanceof Error ? err : new Error(String(err)))\n        }\n      })\n    })\n  } finally {\n    await isolated.cleanup()\n  }\n}\n\nconst CODEX_EXECUTION_POLICY: CodexExecutionPolicy = {\n  sessionPersistence: 'ephemeral',\n  userConfig: false,\n  rules: false,\n  projectInstructions: false,\n  skillInstructions: false,\n  appInstructions: false,\n  toolSuggestions: false,\n  multiAgentInstructions: false,\n  sandbox: 'workspace-write',\n  permissionProfile: 'agent_runtime_reproducible',\n  approvalPolicy: 'never',\n  shellNetwork: false,\n  webSearch: false,\n  serviceTier: 'default',\n  shellEnvironment: 'core-filtered',\n  loginShell: false,\n  credentialsReadable: false,\n  parentRepoRead: false,\n  gitMetadata: false,\n  temporaryDirectory: 'workspace-private',\n  containerSockets: false,\n}\n\nconst FORBIDDEN_CODEX_PROMPT_MARKERS = [\n  '<skills_instructions>',\n  '# AGENTS.md instructions',\n  '<apps_instructions>',\n  '<recommended_plugins>',\n  '<multi_agent_mode>',\n  'You are `/root`, the primary agent',\n] as const\n\nasync function collectCodexExecutionEvidence(opts: {\n  command: string\n  args: string[]\n  cwd: string\n  env: NodeJS.ProcessEnv\n  spawn: NonNullable<RunLocalHarnessOptions['spawn']>\n  controlledConfigSha256: string\n  ambientAuth: string\n  isolatedAuth: string\n  writeProbe: string\n  signal?: AbortSignal\n}): Promise<CodexExecutionEvidence> {\n  assertCodexReproducibleInvocation(opts.command, opts.args)\n  const prompt = opts.args[1]\n  if (opts.args[0] !== 'exec' || typeof prompt !== 'string' || prompt.length === 0) {\n    throw new Error('runLocalHarness: reproducible Codex invocation must be codex exec <prompt>')\n  }\n  const cliVersion = (await runCodexProbe(opts, ['--version'])).stdout.trim()\n  if (!/^codex-cli \\d+\\.\\d+\\.\\d+(?:[-+].+)?$/.test(cliVersion)) {\n    throw new Error(\n      `runLocalHarness: unexpected Codex CLI version output ${JSON.stringify(cliVersion)}`,\n    )\n  }\n\n  const hiddenEnvironmentChecks = [\n    'CODEX_HOME',\n    'CODEX_ACCESS_TOKEN',\n    'CODEX_API_KEY',\n    'OPENAI_API_KEY',\n  ]\n    .map((name) => `test -z \"\\${${name}:-}\"`)\n    .join(' && ')\n  const sandboxProof = [\n    hiddenEnvironmentChecks,\n    'test -r .',\n    'touch \"$1\"',\n    'rm \"$1\"',\n    '! test -r \"$2\"',\n    '! test -r \"$3\"',\n    '! test -r \"$4\"',\n    'touch \"$TMPDIR/.agent-runtime-tmp-proof\"',\n    'rm \"$TMPDIR/.agent-runtime-tmp-proof\"',\n  ].join(' && ')\n  await runCodexProbe(opts, [\n    'sandbox',\n    '-P',\n    CODEX_PERMISSION_PROFILE,\n    '-C',\n    opts.cwd,\n    'sh',\n    '-c',\n    sandboxProof,\n    'sh',\n    opts.writeProbe,\n    opts.ambientAuth,\n    opts.isolatedAuth,\n    join(homedir(), 'code'),\n  ])\n  await runCodexProbe(opts, [\n    'sandbox',\n    '-P',\n    CODEX_PERMISSION_PROFILE,\n    '-C',\n    opts.cwd,\n    'python3',\n    '-c',\n    'import socket,sys\\nfor path in sys.argv[1:]:\\n s=socket.socket(socket.AF_UNIX)\\n try: s.connect(path)\\n except OSError: continue\\n raise SystemExit(1)',\n    '/var/run/docker.sock',\n    '/run/docker.sock',\n    '/var/run/containerd/containerd.sock',\n    '/run/containerd/containerd.sock',\n  ])\n\n  const debugArgs = codexPromptDebugArgs(opts.args, prompt)\n  const renderedPrompt = (await runCodexProbe(opts, debugArgs)).stdout\n  let parsedPrompt: unknown\n  try {\n    parsedPrompt = JSON.parse(renderedPrompt)\n  } catch {\n    throw new Error('runLocalHarness: codex debug prompt-input returned invalid JSON')\n  }\n  if (!Array.isArray(parsedPrompt) || !jsonTextContains(parsedPrompt, prompt)) {\n    throw new Error('runLocalHarness: Codex prompt evidence did not contain the exact task prompt')\n  }\n  const promptContextWithoutTask = JSON.stringify(redactJsonText(parsedPrompt, prompt))\n  for (const marker of FORBIDDEN_CODEX_PROMPT_MARKERS) {\n    if (promptContextWithoutTask.includes(marker)) {\n      throw new Error(`runLocalHarness: Codex prompt evidence contains forbidden marker ${marker}`)\n    }\n  }\n  for (const marker of ['<permissions instructions>', '<environment_context>']) {\n    if (!renderedPrompt.includes(marker)) {\n      throw new Error(`runLocalHarness: Codex prompt evidence is missing required marker ${marker}`)\n    }\n  }\n\n  const nonPromptArgs = [\n    opts.command,\n    ...opts.args.map((arg, index) => (index === 1 ? '<PROMPT>' : arg)),\n  ]\n  return {\n    cliVersion,\n    effectivePromptSha256: sha256(renderedPrompt),\n    nonPromptArgsSha256: sha256(JSON.stringify(nonPromptArgs)),\n    controlledConfigSha256: opts.controlledConfigSha256,\n    policy: { ...CODEX_EXECUTION_POLICY },\n  }\n}\n\nfunction assertCodexReproducibleInvocation(command: string, args: string[]): void {\n  const executable = basename(command)\n  if (executable !== 'codex' && executable !== 'codex.exe') {\n    throw new Error('runLocalHarness: reproducible Codex invocation must use the Codex executable')\n  }\n  const prompt = args[1]\n  const expectedPrefix = ['exec', prompt, ...CODEX_REPRODUCIBLE_ARGS]\n  const hasExactPrefix = expectedPrefix.every((arg, index) => args[index] === arg)\n  const tail = args.slice(expectedPrefix.length)\n  const validModel =\n    tail[0] === '-m' &&\n    typeof tail[1] === 'string' &&\n    tail[1].length > 0 &&\n    !tail[1].startsWith('-')\n  const validReasoning =\n    tail[2] === '-c' &&\n    typeof tail[3] === 'string' &&\n    /^model_reasoning_effort=\"(?:none|minimal|low|medium|high|xhigh)\"$/.test(tail[3])\n  if (!hasExactPrefix || tail.length !== 4 || !validModel || !validReasoning) {\n    throw new Error(\n      'runLocalHarness: reproducible Codex invocation does not match the required isolated argv',\n    )\n  }\n}\n\nfunction codexPromptDebugArgs(execArgs: string[], prompt: string): string[] {\n  const debugArgs = ['debug', 'prompt-input']\n  for (let index = 2; index < execArgs.length; index += 1) {\n    const arg = execArgs[index]\n    const value = execArgs[index + 1]\n    if (arg === '-c' && value !== undefined) {\n      debugArgs.push('-c', value)\n      index += 1\n    } else if (arg === '--disable' && value !== undefined) {\n      debugArgs.push('--disable', value)\n      index += 1\n    } else if (arg === '-m' && value !== undefined) {\n      debugArgs.push('-c', `model=${JSON.stringify(value)}`)\n      index += 1\n    } else if (arg === '-s' && value !== undefined) {\n      debugArgs.push('-c', `sandbox_mode=${JSON.stringify(value)}`)\n      index += 1\n    }\n  }\n  debugArgs.push(prompt)\n  return debugArgs\n}\n\nfunction runCodexProbe(\n  opts: {\n    command: string\n    cwd: string\n    env: NodeJS.ProcessEnv\n    spawn: NonNullable<RunLocalHarnessOptions['spawn']>\n    signal?: AbortSignal\n  },\n  args: string[],\n): Promise<{ stdout: string; stderr: string }> {\n  return new Promise((resolve, reject) => {\n    let child: ChildProcess\n    try {\n      child = opts.spawn(opts.command, args, { cwd: opts.cwd, env: opts.env, stdio: 'pipe' })\n    } catch (err) {\n      reject(err instanceof Error ? err : new Error(String(err)))\n      return\n    }\n    child.stdin?.end()\n    let stdout = ''\n    let stderr = ''\n    let settled = false\n    const timer = setTimeout(() => {\n      if (!child.killed) child.kill('SIGTERM')\n    }, 10_000)\n    timer.unref?.()\n    const onAbort = () => {\n      if (!child.killed) child.kill('SIGTERM')\n    }\n    if (opts.signal) {\n      if (opts.signal.aborted) onAbort()\n      else opts.signal.addEventListener('abort', onAbort, { once: true })\n    }\n    child.stdout?.on('data', (chunk) => {\n      stdout += String(chunk)\n    })\n    child.stderr?.on('data', (chunk) => {\n      stderr += String(chunk)\n    })\n    child.on('error', (err) => {\n      if (settled) return\n      settled = true\n      clearTimeout(timer)\n      opts.signal?.removeEventListener('abort', onAbort)\n      reject(err)\n    })\n    child.on('close', (code, signal) => {\n      if (settled) return\n      settled = true\n      clearTimeout(timer)\n      opts.signal?.removeEventListener('abort', onAbort)\n      if (code !== 0) {\n        const safeStderr = redactCodexHome(stderr, opts.env.CODEX_HOME)\n        reject(\n          new Error(\n            `runLocalHarness: Codex evidence probe failed (exit ${String(code)}, signal ${String(signal)}): ${safeStderr.trim()}`,\n          ),\n        )\n        return\n      }\n      resolve({ stdout, stderr })\n    })\n  })\n}\n\nfunction redactCodexHome(value: string, codexHome: string | undefined): string {\n  return codexHome ? value.replaceAll(codexHome, '<ISOLATED_CODEX_HOME>') : value\n}\n\nfunction sha256(value: string): string {\n  return createHash('sha256').update(value).digest('hex')\n}\n\nfunction jsonTextContains(value: unknown, needle: string): boolean {\n  if (typeof value === 'string') return value.includes(needle)\n  if (Array.isArray(value)) return value.some((item) => jsonTextContains(item, needle))\n  if (isRecord(value)) return Object.values(value).some((item) => jsonTextContains(item, needle))\n  return false\n}\n\nfunction redactJsonText(value: unknown, text: string): unknown {\n  if (typeof value === 'string') return value.replaceAll(text, '')\n  if (Array.isArray(value)) return value.map((item) => redactJsonText(item, text))\n  if (isRecord(value)) {\n    return Object.fromEntries(\n      Object.entries(value).map(([key, item]) => [key, redactJsonText(item, text)]),\n    )\n  }\n  return value\n}\n\nconst CODEX_AUTH_ENV = ['CODEX_ACCESS_TOKEN', 'CODEX_API_KEY', 'OPENAI_API_KEY'] as const\n\nconst CODEX_PERMISSION_PROFILE = 'agent_runtime_reproducible'\n\nasync function isolateCodexHome(\n  baseEnv: NodeJS.ProcessEnv,\n  cwd: string,\n): Promise<{\n  env: NodeJS.ProcessEnv\n  cleanup: () => Promise<void>\n  controlledConfigSha256: string\n  ambientAuth: string\n  isolatedAuth: string\n  writeProbe: string\n}> {\n  const isolatedHome = await mkdtemp(join(tmpdir(), 'agent-runtime-codex-'))\n  await chmod(isolatedHome, 0o700)\n  let workspaceTemp = ''\n  const writeProbe = join(cwd, '.agent-runtime-sandbox-write-proof')\n  const cleanup = async () => {\n    await Promise.all([\n      rm(writeProbe, { force: true }),\n      ...(workspaceTemp ? [rm(workspaceTemp, { recursive: true, force: true })] : []),\n      rm(isolatedHome, { recursive: true, force: true }),\n    ])\n  }\n  try {\n    workspaceTemp = await mkdtemp(join(cwd, '.agent-runtime-tmp-'))\n    await chmod(workspaceTemp, 0o700)\n    const ambientHome = baseEnv.CODEX_HOME?.trim() || join(homedir(), '.codex')\n    const ambientAuth = join(ambientHome, 'auth.json')\n    const isolatedAuth = join(isolatedHome, 'auth.json')\n    let linkedAuth = false\n    try {\n      await access(ambientAuth, constants.R_OK)\n      await symlink(ambientAuth, isolatedAuth)\n      linkedAuth = true\n    } catch (err) {\n      if ((err as NodeJS.ErrnoException).code !== 'ENOENT') throw err\n    }\n    if (!linkedAuth && !CODEX_AUTH_ENV.some((name) => Boolean(baseEnv[name]))) {\n      throw new Error(\n        'runLocalHarness: reproducible Codex mode requires readable auth.json or an API token environment variable',\n      )\n    }\n    const controlledConfig = codexControlledConfig({\n      ambientHome,\n      ambientTmp: baseEnv.TMPDIR?.trim() || tmpdir(),\n      workspaceTemp,\n    })\n    await writeFile(join(isolatedHome, 'config.toml'), controlledConfig, { mode: 0o600 })\n    return {\n      env: { ...baseEnv, CODEX_HOME: isolatedHome },\n      cleanup,\n      controlledConfigSha256: sha256(controlledConfig),\n      ambientAuth,\n      isolatedAuth,\n      writeProbe,\n    }\n  } catch (err) {\n    await cleanup()\n    throw err\n  }\n}\n\nfunction codexControlledConfig(opts: {\n  ambientHome: string\n  ambientTmp: string\n  workspaceTemp: string\n}): string {\n  const home = homedir()\n  const deniedPaths = [\n    opts.ambientHome,\n    opts.ambientTmp,\n    join(home, 'code'),\n    join(home, 'company'),\n    join(home, '.ssh'),\n    join(home, '.aws'),\n    join(home, '.config'),\n    join(home, '.cache'),\n    join(home, '.docker'),\n    join(home, '.kube'),\n    join(home, '.gnupg'),\n    join(home, '.pki'),\n    join(home, '.npmrc'),\n    join(home, '.netrc'),\n    join(home, '.git-credentials'),\n    '/run/docker',\n    '/run/containerd',\n    opts.workspaceTemp,\n  ]\n  const filesystemRules = deniedPaths\n    .map(\n      (path) => `${JSON.stringify(path)} = ${path === opts.workspaceTemp ? '\"write\"' : '\"deny\"'}`,\n    )\n    .join('\\n')\n  return `default_permissions = \"${CODEX_PERMISSION_PROFILE}\"\napproval_policy = \"never\"\nallow_login_shell = false\nservice_tier = \"default\"\n\n[shell_environment_policy]\ninherit = \"core\"\nignore_default_excludes = false\nexclude = [\"^CODEX_HOME$\", \"(?i).*TOKEN.*\", \"(?i).*KEY.*\", \"(?i).*SECRET.*\", \"(?i).*PASSWORD.*\", \"(?i).*CREDENTIAL.*\", \"(?i).*AUTH.*\", \"(?i).*COOKIE.*\"]\nset = { TMPDIR = ${JSON.stringify(opts.workspaceTemp)} }\n\n[permissions.${CODEX_PERMISSION_PROFILE}]\nextends = \":workspace\"\n\n[permissions.${CODEX_PERMISSION_PROFILE}.filesystem]\n${filesystemRules}\n\":slash_tmp\" = \"deny\"\n\n[permissions.${CODEX_PERMISSION_PROFILE}.filesystem.\":workspace_roots\"]\n\".\" = \"write\"\n\n[permissions.${CODEX_PERMISSION_PROFILE}.network]\nenabled = false\n\n[permissions.${CODEX_PERMISSION_PROFILE}.network.unix_sockets]\n\"/var/run/docker.sock\" = \"deny\"\n\"/run/docker.sock\" = \"deny\"\n\"/var/run/containerd/containerd.sock\" = \"deny\"\n\"/run/containerd/containerd.sock\" = \"deny\"\n`\n}\n\n/** Parse and validate the one terminal usage event emitted by `codex exec --json`. */\nexport function parseCodexTokenUsage(stdout: string): CodexTokenUsage {\n  const completed: unknown[] = []\n  for (const [index, line] of stdout.split(/\\r?\\n/).entries()) {\n    if (line.trim().length === 0) continue\n    let event: unknown\n    try {\n      event = JSON.parse(line)\n    } catch {\n      throw new Error(`runLocalHarness: Codex JSONL line ${index + 1} is not valid JSON`)\n    }\n    if (isRecord(event) && event.type === 'turn.completed') completed.push(event)\n  }\n  if (completed.length !== 1) {\n    throw new Error(\n      `runLocalHarness: expected exactly one Codex turn.completed usage event, received ${completed.length}`,\n    )\n  }\n  const event = completed[0]\n  if (!isRecord(event) || !isRecord(event.usage)) {\n    throw new Error('runLocalHarness: Codex turn.completed event is missing usage')\n  }\n  const usage = event.usage\n  const inputTokens = naturalNumber(usage.input_tokens, 'input_tokens')\n  const cachedInputTokens = naturalNumber(usage.cached_input_tokens, 'cached_input_tokens')\n  const outputTokens = naturalNumber(usage.output_tokens, 'output_tokens')\n  const reasoningOutputTokens = naturalNumber(\n    usage.reasoning_output_tokens,\n    'reasoning_output_tokens',\n  )\n  if (cachedInputTokens > inputTokens) {\n    throw new Error('runLocalHarness: Codex cached_input_tokens exceeds input_tokens')\n  }\n  if (reasoningOutputTokens > outputTokens) {\n    throw new Error('runLocalHarness: Codex reasoning_output_tokens exceeds output_tokens')\n  }\n  return { inputTokens, cachedInputTokens, outputTokens, reasoningOutputTokens }\n}\n\nfunction isRecord(value: unknown): value is Record<string, unknown> {\n  return value !== null && typeof value === 'object' && !Array.isArray(value)\n}\n\nfunction naturalNumber(value: unknown, field: string): number {\n  if (typeof value !== 'number' || !Number.isSafeInteger(value) || value < 0) {\n    throw new Error(`runLocalHarness: Codex usage.${field} must be a non-negative safe integer`)\n  }\n  return value\n}\n"],"mappings":";AAmBA,SAA4B,aAAa;AACzC,SAAS,kBAAkB;AAC3B,SAAS,iBAAiB;AAC1B,SAAS,QAAQ,OAAO,SAAS,IAAI,SAAS,iBAAiB;AAC/D,SAAS,SAAS,cAAc;AAChC,SAAS,UAAU,YAAY;AAQ/B,IAAM,uBAAwD;AAAA,EAC5D,MAAM;AAAA,EACN,SAAS;AAAA,EACT,KAAK;AAAA,EACL,QAAQ;AAAA,EACR,MAAM;AAAA,EACN,OAAO;AAAA,EACP,WAAW;AACb;AAEA,SAAS,mBAAmB,iBAA4C;AACtE,QAAM,SAAS,qBAAqB,eAAe;AACnD,MAAI,WAAW,QAAW;AACxB,UAAM,IAAI;AAAA,MACR,yDAAyD,OAAO,eAAe,CAAC;AAAA,IAClF;AAAA,EACF;AACA,SAAO,CAAC,MAAM,2BAA2B,MAAM,GAAG;AACpD;AAUA,IAAM,sBAUF;AAAA,EACF,QAAQ;AAAA,IACN,SAAS;AAAA;AAAA;AAAA,IAGT,WAAW,CAAC,eAAe,CAAC,MAAM,UAAU;AAAA,IAC5C,WAAW,CAAC,UAAU,CAAC,MAAM,KAAK;AAAA,EACpC;AAAA,EACA,OAAO;AAAA,IACL,SAAS;AAAA,IACT,WAAW,CAAC,eAAe,CAAC,QAAQ,UAAU;AAAA,IAC9C,WAAW,CAAC,UAAU,CAAC,MAAM,KAAK;AAAA,IAClC,eAAe;AAAA,EACjB;AAAA,EACA,UAAU;AAAA,IACR,SAAS;AAAA,IACT,WAAW,CAAC,eAAe,CAAC,OAAO,UAAU;AAAA,IAC7C,WAAW,CAAC,UAAU,CAAC,MAAM,KAAK;AAAA,EACpC;AACF;AAiBA,IAAM,0BAA0B;AAAA,EAC9B;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAEA,SAAS,iBACP,SACA,YACA,UAAoC,CAAC,GAC3B;AACV,QAAM,OAAO,oBAAoB,OAAO,EAAE,UAAU,UAAU;AAC9D,MAAI,YAAY,YAAY,QAAQ,4BAA4B;AAC9D,SAAK,KAAK,gCAAgC;AAAA,EAC5C;AACA,MAAI,QAAQ,mBAAmB;AAC7B,QAAI,YAAY,SAAS;AACvB,YAAM,IAAI,MAAM,iEAAiE;AAAA,IACnF;AACA,SAAK,KAAK,GAAG,uBAAuB;AAAA,EACtC;AACA,SAAO;AACT;AAkBO,SAAS,kBACd,SACA,SACA,YACA,UAAoC,CAAC,GAClB;AACnB,QAAM,aAAa,oBAAoB,OAAO;AAC9C,MAAI,CAAC,YAAY;AACf,UAAM,IAAI,MAAM,sCAAsC,OAAO,OAAO,CAAC,EAAE;AAAA,EACzE;AAEA,MAAI,QAAQ,mBAAmB;AAC7B,UAAMA,SAAQ,QAAQ,OAAO;AAC7B,QAAI,OAAOA,WAAU,YAAYA,OAAM,KAAK,EAAE,WAAW,GAAG;AAC1D,YAAM,IAAI,MAAM,qEAAqE;AAAA,IACvF;AACA,QAAI,QAAQ,OAAO,oBAAoB,QAAW;AAChD,YAAM,IAAI,MAAM,6EAA6E;AAAA,IAC/F;AAAA,EACF;AAEA,QAAM,eAAe,QAAQ,QAAQ;AACrC,QAAM,eAAe,QAAQ,QAAQ;AACrC,MACE,iBAAiB,WAChB,CAAC,MAAM,QAAQ,YAAY,KAC1B,aAAa,KAAK,CAAC,gBAAgB,OAAO,gBAAgB,QAAQ,IACpE;AACA,UAAM,IAAI,MAAM,4EAA4E;AAAA,EAC9F;AACA,QAAM,iBAAiB;AAAA,IACrB,GAAI,OAAO,iBAAiB,YAAY,aAAa,KAAK,EAAE,SAAS,IAAI,CAAC,YAAY,IAAI,CAAC;AAAA,IAC3F,GAAI,cAAc,OAAO,CAAC,gBAAgB,YAAY,KAAK,EAAE,SAAS,CAAC,KAAK,CAAC;AAAA,IAC7E;AAAA,EACF;AACA,QAAM,iBAAiB,eAAe,KAAK,MAAM;AAEjD,QAAM,OAAO,iBAAiB,SAAS,gBAAgB,OAAO;AAE9D,QAAM,QAAQ,QAAQ,OAAO;AAC7B,MAAI,OAAO,UAAU,YAAY,MAAM,SAAS,GAAG;AACjD,SAAK,KAAK,GAAG,WAAW,UAAU,KAAK,CAAC;AAAA,EAC1C;AAEA,QAAM,kBAAkB,QAAQ,OAAO;AACvC,MAAI,oBAAoB,UAAa,WAAW,eAAe;AAC7D,SAAK,KAAK,GAAG,WAAW,cAAc,eAAe,CAAC;AAAA,EACxD;AAEA,SAAO,EAAE,SAAS,WAAW,SAAS,KAAK;AAC7C;AA0GA,IAAM,qBAAqB,IAAI,KAAK;AAqBpC,eAAsB,gBACpB,SAC6B;AAC7B,QAAM,EAAE,SAAS,KAAK,WAAW,IAAI;AACrC,MAAI,QAAQ,qBAAqB,YAAY,SAAS;AACpD,UAAM,IAAI,MAAM,+DAA+D;AAAA,EACjF;AACA,MAAI,QAAQ,qBAAqB,QAAQ,aAAa,SAAS;AAC7D,UAAM,IAAI,MAAM,6DAA6D;AAAA,EAC/E;AACA,QAAM,YAAY,QAAQ,aAAa;AACvC,QAAM,YAAY,QAAQ,SAAS;AAEnC,QAAM,aAAa,oBAAoB,OAAO;AAC9C,MAAI,CAAC,YAAY;AACf,UAAM,IAAI,MAAM,oCAAoC,OAAO,OAAO,CAAC,EAAE;AAAA,EACvE;AAEA,QAAM,YAAY,KAAK,IAAI;AAC3B,QAAM,UAAU,QAAQ,YAAY,WAAW,WAAW;AAC1D,QAAM,OAAO,QAAQ,aACjB,CAAC,GAAG,QAAQ,WAAW,IAAI,IAC3B,iBAAiB,SAAS,YAAY,OAAO;AACjD,MAAI,QAAQ,kBAAmB,mCAAkC,SAAS,IAAI;AAE9E,QAAM,UAAU,QAAQ,OAAO,QAAQ;AACvC,QAAM,WAAW,QAAQ,oBACrB,MAAM,iBAAiB,SAAS,GAAG,IACnC;AAAA,IACE,KAAK;AAAA,IACL,SAAS,YAAY;AAAA,IACrB,wBAAwB;AAAA,IACxB,aAAa;AAAA,IACb,cAAc;AAAA,IACd,YAAY;AAAA,EACd;AACJ,QAAM,MAAM,SAAS;AAErB,MAAI;AACF,UAAM,WAAW,QAAQ,oBACrB,MAAM,8BAA8B;AAAA,MAClC;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,OAAO;AAAA,MACP,wBAAwB,SAAS;AAAA,MACjC,aAAa,SAAS;AAAA,MACtB,cAAc,SAAS;AAAA,MACvB,YAAY,SAAS;AAAA,MACrB,GAAI,QAAQ,SAAS,EAAE,QAAQ,QAAQ,OAAO,IAAI,CAAC;AAAA,IACrD,CAAC,IACD;AACJ,WAAO,MAAM,IAAI,QAA4B,CAAC,SAAS,WAAW;AAChE,UAAI;AACJ,UAAI;AACF,gBAAQ,UAAU,SAAS,MAAM,EAAE,KAAK,KAAK,OAAO,OAAO,CAAC;AAAA,MAC9D,SAAS,KAAK;AACZ,eAAO,eAAe,QAAQ,MAAM,IAAI,MAAM,OAAO,GAAG,CAAC,CAAC;AAC1D;AAAA,MACF;AAOA,YAAM,OAAO,IAAI;AAEjB,UAAI,SAAS;AACb,UAAI,SAAS;AACb,UAAI,WAAW;AACf,UAAI,UAAU;AAEd,YAAM,QACJ,YAAY,IACR,WAAW,MAAM;AACf,mBAAW;AACX,YAAI,CAAC,MAAM,OAAQ,OAAM,KAAK,SAAS;AAAA,MACzC,GAAG,SAAS,IACZ;AACN,UAAI,SAAS,OAAQ,MAAiC,UAAU,YAAY;AAC1E;AAAC,QAAC,MAAgC,MAAM;AAAA,MAC1C;AAEA,YAAM,UAAU,MAAM;AACpB,YAAI,CAAC,MAAM,OAAQ,OAAM,KAAK,SAAS;AAAA,MACzC;AACA,UAAI,QAAQ,QAAQ;AAClB,YAAI,QAAQ,OAAO,QAAS,SAAQ;AAAA,YAC/B,SAAQ,OAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAAA,MACvE;AAEA,YAAM,QAAQ,GAAG,QAAQ,CAAC,UAAU;AAClC,kBAAU,OAAO,KAAK;AAAA,MACxB,CAAC;AACD,YAAM,QAAQ,GAAG,QAAQ,CAAC,UAAU;AAClC,kBAAU,OAAO,KAAK;AAAA,MACxB,CAAC;AAED,YAAM,WAAW,CAAC,WAA+B;AAC/C,YAAI,QAAS;AACb,kBAAU;AACV,YAAI,MAAO,cAAa,KAAK;AAC7B,gBAAQ,QAAQ,oBAAoB,SAAS,OAAO;AACpD,gBAAQ,MAAM;AAAA,MAChB;AAEA,YAAM,GAAG,SAAS,CAAC,QAAQ;AACzB,YAAI,QAAS;AACb,kBAAU;AACV,YAAI,MAAO,cAAa,KAAK;AAC7B,gBAAQ,QAAQ,oBAAoB,SAAS,OAAO;AACpD,eAAO,GAAG;AAAA,MACZ,CAAC;AAED,YAAM,GAAG,SAAS,CAAC,MAAM,WAAW;AAClC,YAAI;AACF,gBAAM,QAAQ,QAAQ,oBAAoB,qBAAqB,MAAM,IAAI;AACzE,mBAAS;AAAA,YACP,UAAU;AAAA,YACV;AAAA,YACA,QAAQ,gBAAgB,QAAQ,IAAI,UAAU;AAAA,YAC9C,gBAAgB;AAAA,YAChB,YAAY,KAAK,IAAI,IAAI;AAAA,YACzB;AAAA,YACA,GAAI,QAAQ,EAAE,MAAM,IAAI,CAAC;AAAA,YACzB,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC,CAAC;AAAA,QACH,SAAS,KAAK;AACZ,cAAI,QAAS;AACb,oBAAU;AACV,cAAI,MAAO,cAAa,KAAK;AAC7B,kBAAQ,QAAQ,oBAAoB,SAAS,OAAO;AACpD,iBAAO,eAAe,QAAQ,MAAM,IAAI,MAAM,OAAO,GAAG,CAAC,CAAC;AAAA,QAC5D;AAAA,MACF,CAAC;AAAA,IACH,CAAC;AAAA,EACH,UAAE;AACA,UAAM,SAAS,QAAQ;AAAA,EACzB;AACF;AAEA,IAAM,yBAA+C;AAAA,EACnD,oBAAoB;AAAA,EACpB,YAAY;AAAA,EACZ,OAAO;AAAA,EACP,qBAAqB;AAAA,EACrB,mBAAmB;AAAA,EACnB,iBAAiB;AAAA,EACjB,iBAAiB;AAAA,EACjB,wBAAwB;AAAA,EACxB,SAAS;AAAA,EACT,mBAAmB;AAAA,EACnB,gBAAgB;AAAA,EAChB,cAAc;AAAA,EACd,WAAW;AAAA,EACX,aAAa;AAAA,EACb,kBAAkB;AAAA,EAClB,YAAY;AAAA,EACZ,qBAAqB;AAAA,EACrB,gBAAgB;AAAA,EAChB,aAAa;AAAA,EACb,oBAAoB;AAAA,EACpB,kBAAkB;AACpB;AAEA,IAAM,iCAAiC;AAAA,EACrC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAEA,eAAe,8BAA8B,MAWT;AAClC,oCAAkC,KAAK,SAAS,KAAK,IAAI;AACzD,QAAM,SAAS,KAAK,KAAK,CAAC;AAC1B,MAAI,KAAK,KAAK,CAAC,MAAM,UAAU,OAAO,WAAW,YAAY,OAAO,WAAW,GAAG;AAChF,UAAM,IAAI,MAAM,4EAA4E;AAAA,EAC9F;AACA,QAAM,cAAc,MAAM,cAAc,MAAM,CAAC,WAAW,CAAC,GAAG,OAAO,KAAK;AAC1E,MAAI,CAAC,uCAAuC,KAAK,UAAU,GAAG;AAC5D,UAAM,IAAI;AAAA,MACR,wDAAwD,KAAK,UAAU,UAAU,CAAC;AAAA,IACpF;AAAA,EACF;AAEA,QAAM,0BAA0B;AAAA,IAC9B;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,EACG,IAAI,CAAC,SAAS,eAAe,IAAI,MAAM,EACvC,KAAK,MAAM;AACd,QAAM,eAAe;AAAA,IACnB;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,EAAE,KAAK,MAAM;AACb,QAAM,cAAc,MAAM;AAAA,IACxB;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,KAAK;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,KAAK;AAAA,IACL,KAAK;AAAA,IACL,KAAK;AAAA,IACL,KAAK,QAAQ,GAAG,MAAM;AAAA,EACxB,CAAC;AACD,QAAM,cAAc,MAAM;AAAA,IACxB;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,KAAK;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AAED,QAAM,YAAY,qBAAqB,KAAK,MAAM,MAAM;AACxD,QAAM,kBAAkB,MAAM,cAAc,MAAM,SAAS,GAAG;AAC9D,MAAI;AACJ,MAAI;AACF,mBAAe,KAAK,MAAM,cAAc;AAAA,EAC1C,QAAQ;AACN,UAAM,IAAI,MAAM,iEAAiE;AAAA,EACnF;AACA,MAAI,CAAC,MAAM,QAAQ,YAAY,KAAK,CAAC,iBAAiB,cAAc,MAAM,GAAG;AAC3E,UAAM,IAAI,MAAM,8EAA8E;AAAA,EAChG;AACA,QAAM,2BAA2B,KAAK,UAAU,eAAe,cAAc,MAAM,CAAC;AACpF,aAAW,UAAU,gCAAgC;AACnD,QAAI,yBAAyB,SAAS,MAAM,GAAG;AAC7C,YAAM,IAAI,MAAM,oEAAoE,MAAM,EAAE;AAAA,IAC9F;AAAA,EACF;AACA,aAAW,UAAU,CAAC,8BAA8B,uBAAuB,GAAG;AAC5E,QAAI,CAAC,eAAe,SAAS,MAAM,GAAG;AACpC,YAAM,IAAI,MAAM,qEAAqE,MAAM,EAAE;AAAA,IAC/F;AAAA,EACF;AAEA,QAAM,gBAAgB;AAAA,IACpB,KAAK;AAAA,IACL,GAAG,KAAK,KAAK,IAAI,CAAC,KAAK,UAAW,UAAU,IAAI,aAAa,GAAI;AAAA,EACnE;AACA,SAAO;AAAA,IACL;AAAA,IACA,uBAAuB,OAAO,cAAc;AAAA,IAC5C,qBAAqB,OAAO,KAAK,UAAU,aAAa,CAAC;AAAA,IACzD,wBAAwB,KAAK;AAAA,IAC7B,QAAQ,EAAE,GAAG,uBAAuB;AAAA,EACtC;AACF;AAEA,SAAS,kCAAkC,SAAiB,MAAsB;AAChF,QAAM,aAAa,SAAS,OAAO;AACnC,MAAI,eAAe,WAAW,eAAe,aAAa;AACxD,UAAM,IAAI,MAAM,8EAA8E;AAAA,EAChG;AACA,QAAM,SAAS,KAAK,CAAC;AACrB,QAAM,iBAAiB,CAAC,QAAQ,QAAQ,GAAG,uBAAuB;AAClE,QAAM,iBAAiB,eAAe,MAAM,CAAC,KAAK,UAAU,KAAK,KAAK,MAAM,GAAG;AAC/E,QAAM,OAAO,KAAK,MAAM,eAAe,MAAM;AAC7C,QAAM,aACJ,KAAK,CAAC,MAAM,QACZ,OAAO,KAAK,CAAC,MAAM,YACnB,KAAK,CAAC,EAAE,SAAS,KACjB,CAAC,KAAK,CAAC,EAAE,WAAW,GAAG;AACzB,QAAM,iBACJ,KAAK,CAAC,MAAM,QACZ,OAAO,KAAK,CAAC,MAAM,YACnB,oEAAoE,KAAK,KAAK,CAAC,CAAC;AAClF,MAAI,CAAC,kBAAkB,KAAK,WAAW,KAAK,CAAC,cAAc,CAAC,gBAAgB;AAC1E,UAAM,IAAI;AAAA,MACR;AAAA,IACF;AAAA,EACF;AACF;AAEA,SAAS,qBAAqB,UAAoB,QAA0B;AAC1E,QAAM,YAAY,CAAC,SAAS,cAAc;AAC1C,WAAS,QAAQ,GAAG,QAAQ,SAAS,QAAQ,SAAS,GAAG;AACvD,UAAM,MAAM,SAAS,KAAK;AAC1B,UAAM,QAAQ,SAAS,QAAQ,CAAC;AAChC,QAAI,QAAQ,QAAQ,UAAU,QAAW;AACvC,gBAAU,KAAK,MAAM,KAAK;AAC1B,eAAS;AAAA,IACX,WAAW,QAAQ,eAAe,UAAU,QAAW;AACrD,gBAAU,KAAK,aAAa,KAAK;AACjC,eAAS;AAAA,IACX,WAAW,QAAQ,QAAQ,UAAU,QAAW;AAC9C,gBAAU,KAAK,MAAM,SAAS,KAAK,UAAU,KAAK,CAAC,EAAE;AACrD,eAAS;AAAA,IACX,WAAW,QAAQ,QAAQ,UAAU,QAAW;AAC9C,gBAAU,KAAK,MAAM,gBAAgB,KAAK,UAAU,KAAK,CAAC,EAAE;AAC5D,eAAS;AAAA,IACX;AAAA,EACF;AACA,YAAU,KAAK,MAAM;AACrB,SAAO;AACT;AAEA,SAAS,cACP,MAOA,MAC6C;AAC7C,SAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,QAAI;AACJ,QAAI;AACF,cAAQ,KAAK,MAAM,KAAK,SAAS,MAAM,EAAE,KAAK,KAAK,KAAK,KAAK,KAAK,KAAK,OAAO,OAAO,CAAC;AAAA,IACxF,SAAS,KAAK;AACZ,aAAO,eAAe,QAAQ,MAAM,IAAI,MAAM,OAAO,GAAG,CAAC,CAAC;AAC1D;AAAA,IACF;AACA,UAAM,OAAO,IAAI;AACjB,QAAI,SAAS;AACb,QAAI,SAAS;AACb,QAAI,UAAU;AACd,UAAM,QAAQ,WAAW,MAAM;AAC7B,UAAI,CAAC,MAAM,OAAQ,OAAM,KAAK,SAAS;AAAA,IACzC,GAAG,GAAM;AACT,UAAM,QAAQ;AACd,UAAM,UAAU,MAAM;AACpB,UAAI,CAAC,MAAM,OAAQ,OAAM,KAAK,SAAS;AAAA,IACzC;AACA,QAAI,KAAK,QAAQ;AACf,UAAI,KAAK,OAAO,QAAS,SAAQ;AAAA,UAC5B,MAAK,OAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAAA,IACpE;AACA,UAAM,QAAQ,GAAG,QAAQ,CAAC,UAAU;AAClC,gBAAU,OAAO,KAAK;AAAA,IACxB,CAAC;AACD,UAAM,QAAQ,GAAG,QAAQ,CAAC,UAAU;AAClC,gBAAU,OAAO,KAAK;AAAA,IACxB,CAAC;AACD,UAAM,GAAG,SAAS,CAAC,QAAQ;AACzB,UAAI,QAAS;AACb,gBAAU;AACV,mBAAa,KAAK;AAClB,WAAK,QAAQ,oBAAoB,SAAS,OAAO;AACjD,aAAO,GAAG;AAAA,IACZ,CAAC;AACD,UAAM,GAAG,SAAS,CAAC,MAAM,WAAW;AAClC,UAAI,QAAS;AACb,gBAAU;AACV,mBAAa,KAAK;AAClB,WAAK,QAAQ,oBAAoB,SAAS,OAAO;AACjD,UAAI,SAAS,GAAG;AACd,cAAM,aAAa,gBAAgB,QAAQ,KAAK,IAAI,UAAU;AAC9D;AAAA,UACE,IAAI;AAAA,YACF,sDAAsD,OAAO,IAAI,CAAC,YAAY,OAAO,MAAM,CAAC,MAAM,WAAW,KAAK,CAAC;AAAA,UACrH;AAAA,QACF;AACA;AAAA,MACF;AACA,cAAQ,EAAE,QAAQ,OAAO,CAAC;AAAA,IAC5B,CAAC;AAAA,EACH,CAAC;AACH;AAEA,SAAS,gBAAgB,OAAe,WAAuC;AAC7E,SAAO,YAAY,MAAM,WAAW,WAAW,uBAAuB,IAAI;AAC5E;AAEA,SAAS,OAAO,OAAuB;AACrC,SAAO,WAAW,QAAQ,EAAE,OAAO,KAAK,EAAE,OAAO,KAAK;AACxD;AAEA,SAAS,iBAAiB,OAAgB,QAAyB;AACjE,MAAI,OAAO,UAAU,SAAU,QAAO,MAAM,SAAS,MAAM;AAC3D,MAAI,MAAM,QAAQ,KAAK,EAAG,QAAO,MAAM,KAAK,CAAC,SAAS,iBAAiB,MAAM,MAAM,CAAC;AACpF,MAAI,SAAS,KAAK,EAAG,QAAO,OAAO,OAAO,KAAK,EAAE,KAAK,CAAC,SAAS,iBAAiB,MAAM,MAAM,CAAC;AAC9F,SAAO;AACT;AAEA,SAAS,eAAe,OAAgB,MAAuB;AAC7D,MAAI,OAAO,UAAU,SAAU,QAAO,MAAM,WAAW,MAAM,EAAE;AAC/D,MAAI,MAAM,QAAQ,KAAK,EAAG,QAAO,MAAM,IAAI,CAAC,SAAS,eAAe,MAAM,IAAI,CAAC;AAC/E,MAAI,SAAS,KAAK,GAAG;AACnB,WAAO,OAAO;AAAA,MACZ,OAAO,QAAQ,KAAK,EAAE,IAAI,CAAC,CAAC,KAAK,IAAI,MAAM,CAAC,KAAK,eAAe,MAAM,IAAI,CAAC,CAAC;AAAA,IAC9E;AAAA,EACF;AACA,SAAO;AACT;AAEA,IAAM,iBAAiB,CAAC,sBAAsB,iBAAiB,gBAAgB;AAE/E,IAAM,2BAA2B;AAEjC,eAAe,iBACb,SACA,KAQC;AACD,QAAM,eAAe,MAAM,QAAQ,KAAK,OAAO,GAAG,sBAAsB,CAAC;AACzE,QAAM,MAAM,cAAc,GAAK;AAC/B,MAAI,gBAAgB;AACpB,QAAM,aAAa,KAAK,KAAK,oCAAoC;AACjE,QAAM,UAAU,YAAY;AAC1B,UAAM,QAAQ,IAAI;AAAA,MAChB,GAAG,YAAY,EAAE,OAAO,KAAK,CAAC;AAAA,MAC9B,GAAI,gBAAgB,CAAC,GAAG,eAAe,EAAE,WAAW,MAAM,OAAO,KAAK,CAAC,CAAC,IAAI,CAAC;AAAA,MAC7E,GAAG,cAAc,EAAE,WAAW,MAAM,OAAO,KAAK,CAAC;AAAA,IACnD,CAAC;AAAA,EACH;AACA,MAAI;AACF,oBAAgB,MAAM,QAAQ,KAAK,KAAK,qBAAqB,CAAC;AAC9D,UAAM,MAAM,eAAe,GAAK;AAChC,UAAM,cAAc,QAAQ,YAAY,KAAK,KAAK,KAAK,QAAQ,GAAG,QAAQ;AAC1E,UAAM,cAAc,KAAK,aAAa,WAAW;AACjD,UAAM,eAAe,KAAK,cAAc,WAAW;AACnD,QAAI,aAAa;AACjB,QAAI;AACF,YAAM,OAAO,aAAa,UAAU,IAAI;AACxC,YAAM,QAAQ,aAAa,YAAY;AACvC,mBAAa;AAAA,IACf,SAAS,KAAK;AACZ,UAAK,IAA8B,SAAS,SAAU,OAAM;AAAA,IAC9D;AACA,QAAI,CAAC,cAAc,CAAC,eAAe,KAAK,CAAC,SAAS,QAAQ,QAAQ,IAAI,CAAC,CAAC,GAAG;AACzE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,UAAM,mBAAmB,sBAAsB;AAAA,MAC7C;AAAA,MACA,YAAY,QAAQ,QAAQ,KAAK,KAAK,OAAO;AAAA,MAC7C;AAAA,IACF,CAAC;AACD,UAAM,UAAU,KAAK,cAAc,aAAa,GAAG,kBAAkB,EAAE,MAAM,IAAM,CAAC;AACpF,WAAO;AAAA,MACL,KAAK,EAAE,GAAG,SAAS,YAAY,aAAa;AAAA,MAC5C;AAAA,MACA,wBAAwB,OAAO,gBAAgB;AAAA,MAC/C;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF,SAAS,KAAK;AACZ,UAAM,QAAQ;AACd,UAAM;AAAA,EACR;AACF;AAEA,SAAS,sBAAsB,MAIpB;AACT,QAAM,OAAO,QAAQ;AACrB,QAAM,cAAc;AAAA,IAClB,KAAK;AAAA,IACL,KAAK;AAAA,IACL,KAAK,MAAM,MAAM;AAAA,IACjB,KAAK,MAAM,SAAS;AAAA,IACpB,KAAK,MAAM,MAAM;AAAA,IACjB,KAAK,MAAM,MAAM;AAAA,IACjB,KAAK,MAAM,SAAS;AAAA,IACpB,KAAK,MAAM,QAAQ;AAAA,IACnB,KAAK,MAAM,SAAS;AAAA,IACpB,KAAK,MAAM,OAAO;AAAA,IAClB,KAAK,MAAM,QAAQ;AAAA,IACnB,KAAK,MAAM,MAAM;AAAA,IACjB,KAAK,MAAM,QAAQ;AAAA,IACnB,KAAK,MAAM,QAAQ;AAAA,IACnB,KAAK,MAAM,kBAAkB;AAAA,IAC7B;AAAA,IACA;AAAA,IACA,KAAK;AAAA,EACP;AACA,QAAM,kBAAkB,YACrB;AAAA,IACC,CAAC,SAAS,GAAG,KAAK,UAAU,IAAI,CAAC,MAAM,SAAS,KAAK,gBAAgB,YAAY,QAAQ;AAAA,EAC3F,EACC,KAAK,IAAI;AACZ,SAAO,0BAA0B,wBAAwB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mBASxC,KAAK,UAAU,KAAK,aAAa,CAAC;AAAA;AAAA,eAEtC,wBAAwB;AAAA;AAAA;AAAA,eAGxB,wBAAwB;AAAA,EACrC,eAAe;AAAA;AAAA;AAAA,eAGF,wBAAwB;AAAA;AAAA;AAAA,eAGxB,wBAAwB;AAAA;AAAA;AAAA,eAGxB,wBAAwB;AAAA;AAAA;AAAA;AAAA;AAAA;AAMvC;AAGO,SAAS,qBAAqB,QAAiC;AACpE,QAAM,YAAuB,CAAC;AAC9B,aAAW,CAAC,OAAO,IAAI,KAAK,OAAO,MAAM,OAAO,EAAE,QAAQ,GAAG;AAC3D,QAAI,KAAK,KAAK,EAAE,WAAW,EAAG;AAC9B,QAAIC;AACJ,QAAI;AACF,MAAAA,SAAQ,KAAK,MAAM,IAAI;AAAA,IACzB,QAAQ;AACN,YAAM,IAAI,MAAM,qCAAqC,QAAQ,CAAC,oBAAoB;AAAA,IACpF;AACA,QAAI,SAASA,MAAK,KAAKA,OAAM,SAAS,iBAAkB,WAAU,KAAKA,MAAK;AAAA,EAC9E;AACA,MAAI,UAAU,WAAW,GAAG;AAC1B,UAAM,IAAI;AAAA,MACR,oFAAoF,UAAU,MAAM;AAAA,IACtG;AAAA,EACF;AACA,QAAM,QAAQ,UAAU,CAAC;AACzB,MAAI,CAAC,SAAS,KAAK,KAAK,CAAC,SAAS,MAAM,KAAK,GAAG;AAC9C,UAAM,IAAI,MAAM,8DAA8D;AAAA,EAChF;AACA,QAAM,QAAQ,MAAM;AACpB,QAAM,cAAc,cAAc,MAAM,cAAc,cAAc;AACpE,QAAM,oBAAoB,cAAc,MAAM,qBAAqB,qBAAqB;AACxF,QAAM,eAAe,cAAc,MAAM,eAAe,eAAe;AACvE,QAAM,wBAAwB;AAAA,IAC5B,MAAM;AAAA,IACN;AAAA,EACF;AACA,MAAI,oBAAoB,aAAa;AACnC,UAAM,IAAI,MAAM,iEAAiE;AAAA,EACnF;AACA,MAAI,wBAAwB,cAAc;AACxC,UAAM,IAAI,MAAM,sEAAsE;AAAA,EACxF;AACA,SAAO,EAAE,aAAa,mBAAmB,cAAc,sBAAsB;AAC/E;AAEA,SAAS,SAAS,OAAkD;AAClE,SAAO,UAAU,QAAQ,OAAO,UAAU,YAAY,CAAC,MAAM,QAAQ,KAAK;AAC5E;AAEA,SAAS,cAAc,OAAgB,OAAuB;AAC5D,MAAI,OAAO,UAAU,YAAY,CAAC,OAAO,cAAc,KAAK,KAAK,QAAQ,GAAG;AAC1E,UAAM,IAAI,MAAM,gCAAgC,KAAK,sCAAsC;AAAA,EAC7F;AACA,SAAO;AACT;","names":["model","event"]}

package/dist/{chunk-MRYPFDHJ.js → chunk-RTSZNYXM.js}

@@ -5,13 +5,13 @@ import {
   definePersona,
   runPersonified,
   worktreeFanout
-} from "./chunk-OO4EK3JB.js";
+} from "./chunk-D336OZHK.js";
 import {
   runAnalystLoop
 } from "./chunk-ZQZX77MM.js";
 import {
   createExecutorRegistry
-} from "./chunk-FHZHADIM.js";
+} from "./chunk-URMJTSXB.js";
 import {
   ConfigError
 } from "./chunk-YEJR7IXO.js";
@@ -196,4 +196,4 @@ export {
   runLoopRunnerCli,
   parseLoopRunnerArgv
 };
-//# sourceMappingURL=chunk-MRYPFDHJ.js.map
+//# sourceMappingURL=chunk-RTSZNYXM.js.map

package/dist/{chunk-H2D3DCEB.js → chunk-SJHFJOCY.js}

@@ -1,6 +1,6 @@
 import {
   agenticGenerator
-} from "./chunk-RTVVH2KL.js";
+} from "./chunk-JEP2WM24.js";
 import {
   assertModelAllowed
 } from "./chunk-ISTDY47H.js";
@@ -532,4 +532,4 @@ export {
   applyImprovementWinnerToProfile,
   improve
 };
-//# sourceMappingURL=chunk-H2D3DCEB.js.map
+//# sourceMappingURL=chunk-SJHFJOCY.js.map

package/dist/{chunk-FHZHADIM.js → chunk-URMJTSXB.js}

@@ -9,7 +9,7 @@ import {
   InMemoryResultBlobStore,
   InMemorySpawnJournal,
   contentAddress
-} from "./chunk-3D2RHC4K.js";
+} from "./chunk-WIPGQ4GT.js";
 import {
   providerAsExecutor,
   resolveAgentEnvironmentProvider
@@ -33,7 +33,7 @@ import {
 import {
   harnessInvocation,
   runLocalHarness
-} from "./chunk-PX6SXX3M.js";
+} from "./chunk-PJNGPWTA.js";
 import {
   AgentEvalError,
   ConfigError,
@@ -620,6 +620,7 @@ function clampSpend(spend, budget) {
       output: Math.floor(spend.tokens.output * ratio)
     } : spend.tokens,
     usd: budget.maxUsd === void 0 ? spend.usd : Math.min(spend.usd, budget.maxUsd),
+    ...spend.usdKnown === false ? { usdKnown: false } : {},
     ms: spend.ms
   };
 }
@@ -1980,13 +1981,15 @@ async function runWorktreeHarness(opts) {
     const { command, args } = harnessInvocation(opts.harness, opts.profile, opts.taskPrompt, {
       // This helper created the candidate worktree above; autonomous Claude
       // edits are permitted only inside that isolated checkout.
-      dangerouslySkipPermissions: opts.harness === "claude"
+      dangerouslySkipPermissions: opts.harness === "claude",
+      ...opts.codexReproducible ? { codexReproducible: true } : {}
     });
     const harnessResult = await runHarness({
       harness: opts.harness,
       cwd: worktree.path,
       taskPrompt: opts.taskPrompt,
       invocation: { command, args },
+      ...opts.codexReproducible ? { codexReproducible: true } : {},
       ...opts.harnessTimeoutMs !== void 0 ? { timeoutMs: opts.harnessTimeoutMs } : {},
       ...opts.signal ? { signal: opts.signal } : {}
     });
@@ -2014,7 +2017,15 @@ async function runWorktreeHarness(opts) {
         killedBySignal: harnessResult.killedBySignal,
         durationMs: harnessResult.durationMs,
         stdout: harnessResult.stdout,
-        stderr: harnessResult.stderr
+        stderr: harnessResult.stderr,
+        ...harnessResult.usage ? { usage: harnessResult.usage } : {},
+        ...harnessResult.evidence ? {
+          cliVersion: harnessResult.evidence.cliVersion,
+          effectivePromptSha256: harnessResult.evidence.effectivePromptSha256,
+          nonPromptArgsSha256: harnessResult.evidence.nonPromptArgsSha256,
+          controlledConfigSha256: harnessResult.evidence.controlledConfigSha256,
+          executionPolicy: harnessResult.evidence.policy
+        } : {}
       },
       ...checks ? { checks } : {}
     };
@@ -2082,9 +2093,17 @@ function createWorktreeCliExecutor(options) {
   if (typeof options.taskPrompt !== "string" || options.taskPrompt.length === 0) {
     throw new ValidationError("createWorktreeCliExecutor: taskPrompt required");
   }
+  if (options.codexReproducible && options.harness !== "codex") {
+    throw new ValidationError(
+      'createWorktreeCliExecutor: codexReproducible requires harness "codex"'
+    );
+  }
+  if (options.codexReproducible && options.budgetExempt === true) {
+    throw new ValidationError("createWorktreeCliExecutor: codexReproducible cannot be budgetExempt");
+  }
   const runId = options.runId ?? randomUUID();
   const controller = new AbortController();
-  const budgetExempt = options.budgetExempt ?? true;
+  const budgetExempt = options.budgetExempt ?? !options.codexReproducible;
   let run;
   let artifact;
   return {
@@ -2103,6 +2122,7 @@ function createWorktreeCliExecutor(options) {
         ...options.testCmd !== void 0 ? { testCmd: options.testCmd } : {},
         ...options.typecheckCmd !== void 0 ? { typecheckCmd: options.typecheckCmd } : {},
         ...options.harnessTimeoutMs !== void 0 ? { harnessTimeoutMs: options.harnessTimeoutMs } : {},
+        ...options.codexReproducible ? { codexReproducible: true } : {},
         ...options.checkTimeoutMs !== void 0 ? { checkTimeoutMs: options.checkTimeoutMs } : {},
         ...options.checkOutputCap !== void 0 ? { checkOutputCap: options.checkOutputCap } : {},
         ...linked ? { signal: linked } : {},
@@ -2110,13 +2130,20 @@ function createWorktreeCliExecutor(options) {
         ...options.runHarness ? { runHarness: options.runHarness } : {},
         ...options.runCommand ? { runCommand: options.runCommand } : {}
       });
+      const usage = run.result.harness.usage;
+      if (!budgetExempt && !usage) {
+        const completed = run;
+        run = void 0;
+        await completed.cleanup();
+        throw new ValidationError(
+          "createWorktreeCliExecutor: metered harness run returned no token usage"
+        );
+      }
       const spent = {
         iterations: 1,
-        // The worktree-harness core surfaces no token/usd usage, so tokens/usd are a genuine zero
-        // (NOT a fabricated cost). When budgetExempt is true the pool ignores this spend entirely;
-        // when explicitly false the scope debits exactly this captured spend — the real iteration.
-        tokens: zeroTokenUsage(),
+        tokens: usage ? { input: usage.inputTokens, output: usage.outputTokens } : { input: 0, output: 0 },
         usd: 0,
+        ...usage ? { usdKnown: false } : {},
         ms: Date.now() - started
       };
       artifact = { outRef: contentAddress(run.result), out: run.result, spent };
@@ -3090,6 +3117,7 @@ var cliWorktreeExecutor = (spec, ctx) => {
     ...seam.runId ? { runId: seam.runId } : {},
     ...seam.baseRef ? { baseRef: seam.baseRef } : {},
     ...seam.harnessTimeoutMs !== void 0 ? { harnessTimeoutMs: seam.harnessTimeoutMs } : {},
+    ...seam.codexReproducible ? { codexReproducible: true } : {},
     ...seam.testCmd !== void 0 ? { testCmd: seam.testCmd } : {},
     ...seam.typecheckCmd !== void 0 ? { typecheckCmd: seam.typecheckCmd } : {},
     ...seam.checkTimeoutMs !== void 0 ? { checkTimeoutMs: seam.checkTimeoutMs } : {},
@@ -3332,6 +3360,7 @@ function sumSpend(settled) {
     total.tokens.input += ev.spent.tokens.input;
     total.tokens.output += ev.spent.tokens.output;
     total.usd += ev.spent.usd;
+    if (ev.spent.usdKnown === false) total.usdKnown = false;
     total.ms += ev.spent.ms;
   }
   return total;
@@ -3344,6 +3373,7 @@ function sumMetered(events) {
     total.tokens.input += ev.spend.tokens.input;
     total.tokens.output += ev.spend.tokens.output;
     total.usd += ev.spend.usd;
+    if (ev.spend.usdKnown === false) total.usdKnown = false;
     total.ms += ev.spend.ms;
   }
   return total;
@@ -3488,6 +3518,11 @@ function createBudgetPool(root, now = Date.now) {
     }
     open.delete(ticket.id);
     const { tokens: rTokens, usd: rUsd, iterations: rIterations } = ticket.reserved;
+    if (usdCapped && spent.usdKnown === false) {
+      throw new Error(
+        `budget pool: ticket ${ticket.id} reported unknown dollar cost under a dollar-capped budget`
+      );
+    }
     const spentTokens = totalTokens(spent.tokens);
     if (spentTokens > rTokens) {
       throw new Error(
@@ -3517,6 +3552,11 @@ function createBudgetPool(root, now = Date.now) {
     }
   }
   function observe(spend) {
+    if (usdCapped && spend.usdKnown === false) {
+      throw new Error(
+        "budget pool: cannot observe unknown dollar cost under a dollar-capped budget"
+      );
+    }
     const tokens = totalTokens(spend.tokens);
     freeTokens -= tokens;
     committedTokens += tokens;
@@ -3741,6 +3781,7 @@ function accumulate(a, b) {
   a.tokens.input += b.tokens.input;
   a.tokens.output += b.tokens.output;
   a.usd += b.usd;
+  if (b.usdKnown === false) a.usdKnown = false;
   a.ms += b.ms;
 }
 function addSpend(a, b) {
@@ -3748,6 +3789,7 @@ function addSpend(a, b) {
     iterations: a.iterations + b.iterations,
     tokens: { input: a.tokens.input + b.tokens.input, output: a.tokens.output + b.tokens.output },
     usd: a.usd + b.usd,
+    ...a.usdKnown === false || b.usdKnown === false ? { usdKnown: false } : {},
     ms: a.ms + b.ms
   };
 }
@@ -6571,4 +6613,4 @@ export {
   workerFromBackend,
   supervise
 };
-//# sourceMappingURL=chunk-FHZHADIM.js.map
+//# sourceMappingURL=chunk-URMJTSXB.js.map