Merge pull request #2302 from thedotmack/codex/remove-agent-pool-timeout

[codex] Remove agent pool timeout data loss

src/npx-cli/commands/install.ts

@@ -7,6 +7,7 @@ import { homedir } from 'os';
 import { dirname, join } from 'path';
 import { SettingsDefaultsManager, type SettingsDefaults } from '../../shared/SettingsDefaultsManager.js';
 import { USER_SETTINGS_PATH } from '../../shared/paths.js';
+import { loadClaudeMemEnv, saveClaudeMemEnv } from '../../shared/EnvManager.js';
 import { ensureWorkerStarted, type WorkerStartResult } from '../../services/worker-spawner.js';
 import {
   ensureBun,
@@ -588,13 +589,143 @@ function mergeSettings(updates: Record<string, string>): boolean {
 }
 
 type ProviderId = 'claude' | 'gemini' | 'openrouter';
+type ClaudeAccessMode = 'subscription' | 'api-key';
+type ClaudeApiMode = 'direct' | 'gateway';
+
+function readRawStoredAuthMethod(): 'subscription' | 'api-key' | 'gateway' | undefined {
+  try {
+    if (!existsSync(USER_SETTINGS_PATH)) return undefined;
+    const raw = JSON.parse(readFileSync(USER_SETTINGS_PATH, 'utf-8')) as Record<string, unknown>;
+    const flat = (raw.env && typeof raw.env === 'object' ? raw.env : raw) as Record<string, unknown>;
+    const value = flat.CLAUDE_MEM_CLAUDE_AUTH_METHOD;
+    if (value === 'subscription' || value === 'api-key' || value === 'gateway') return value;
+    return undefined;
+  } catch {
+    return undefined;
+  }
+}
+
+function resolveClaudeAuthMethod(): 'subscription' | 'api-key' | 'gateway' {
+  const stored = readRawStoredAuthMethod();
+  if (stored) return stored;
+  const env = loadClaudeMemEnv();
+  if (env.ANTHROPIC_BASE_URL?.trim()) return 'gateway';
+  if (env.ANTHROPIC_API_KEY?.trim()) return 'api-key';
+  return 'subscription';
+}
 
 async function promptProvider(options: InstallOptions): Promise<ProviderId> {
   const initialProvider = (getSetting('CLAUDE_MEM_PROVIDER') as ProviderId) || 'claude';
 
-  const persistClaudeProvider = () => {
-    const wrote = mergeSettings({ CLAUDE_MEM_PROVIDER: 'claude' });
-    if (wrote) log.info('Saved provider=claude to ~/.claude-mem/settings.json');
+  const persistClaudeProvider = (authMethod?: 'subscription' | 'api-key' | 'gateway') => {
+    const resolvedAuthMethod = authMethod ?? resolveClaudeAuthMethod();
+    const wrote = mergeSettings({
+      CLAUDE_MEM_PROVIDER: 'claude',
+      CLAUDE_MEM_CLAUDE_AUTH_METHOD: resolvedAuthMethod,
+    });
+    if (wrote) log.info('Saved Claude Agent SDK configuration to ~/.claude-mem/settings.json');
+  };
+
+  const useSubscriptionAuth = () => {
+    persistClaudeProvider('subscription');
+    saveClaudeMemEnv({
+      ANTHROPIC_API_KEY: '',
+      ANTHROPIC_BASE_URL: '',
+      ANTHROPIC_AUTH_TOKEN: '',
+    });
+    log.info('Configured claude-mem to use your logged-in Claude SDK account.');
+  };
+
+  const configureDirectApiKey = async (): Promise<void> => {
+    const existing = loadClaudeMemEnv().ANTHROPIC_API_KEY || '';
+    if (existing.trim().length > 0) {
+      const choice = await p.select<'keep' | 'replace'>({
+        message: 'An Anthropic API key is already configured. Keep it or enter a new one?',
+        options: [
+          { value: 'keep', label: 'Keep existing key' },
+          { value: 'replace', label: 'Enter a new key (rotate)' },
+        ],
+        initialValue: 'keep',
+      });
+      if (p.isCancel(choice)) {
+        log.warn('API key prompt cancelled — leaving existing configuration untouched.');
+        return;
+      }
+      if (choice === 'keep') {
+        saveClaudeMemEnv({
+          ANTHROPIC_API_KEY: existing.trim(),
+          ANTHROPIC_BASE_URL: '',
+          ANTHROPIC_AUTH_TOKEN: '',
+        });
+        persistClaudeProvider('api-key');
+        return;
+      }
+    }
+
+    const apiKeyResult = await p.password({
+      message: 'Paste your Anthropic API key:',
+      mask: '*',
+      validate: (v?: string) => (!v || v.trim().length === 0) ? 'API key required' : undefined,
+    });
+
+    if (p.isCancel(apiKeyResult)) {
+      log.warn('API key prompt cancelled — leaving existing configuration untouched.');
+      return;
+    }
+
+    saveClaudeMemEnv({
+      ANTHROPIC_API_KEY: String(apiKeyResult).trim(),
+      ANTHROPIC_BASE_URL: '',
+      ANTHROPIC_AUTH_TOKEN: '',
+    });
+    persistClaudeProvider('api-key');
+    log.info('Saved Anthropic API key for the Claude Agent SDK path.');
+  };
+
+  const configureGateway = async (): Promise<void> => {
+    const existing = loadClaudeMemEnv();
+    const baseUrlResult = await p.text({
+      message: 'Gateway URL:',
+      placeholder: existing.ANTHROPIC_BASE_URL || 'http://localhost:4000',
+      defaultValue: existing.ANTHROPIC_BASE_URL || '',
+      validate: (v?: string) => {
+        const value = v?.trim() ?? '';
+        if (!value) return 'Gateway URL required';
+        try {
+          new URL(value);
+          return undefined;
+        } catch {
+          return 'Enter a valid URL, for example http://localhost:4000';
+        }
+      },
+    });
+
+    if (p.isCancel(baseUrlResult)) {
+      log.warn('Gateway setup cancelled — leaving existing configuration untouched.');
+      return;
+    }
+
+    const tokenResult = await p.password({
+      message: 'Gateway key/token (leave blank to keep current token, or type a new one):',
+      mask: '*',
+    });
+
+    const tokenCancelled = p.isCancel(tokenResult);
+    const tokenInput = tokenCancelled ? '' : String(tokenResult).trim();
+    const env: Record<string, string> = {
+      ANTHROPIC_API_KEY: '',
+      ANTHROPIC_BASE_URL: String(baseUrlResult).trim(),
+    };
+    if (!tokenCancelled && tokenInput.length > 0) {
+      env.ANTHROPIC_AUTH_TOKEN = tokenInput;
+    }
+    saveClaudeMemEnv(env);
+    persistClaudeProvider('gateway');
+    if (tokenCancelled || tokenInput.length === 0) {
+      log.info('Gateway URL saved; existing gateway token preserved.');
+    } else {
+      log.info('Configured Claude Agent SDK gateway in ~/.claude-mem/.env.');
+    }
   };
 
   if (!isInteractive) {
@@ -611,29 +742,72 @@ async function promptProvider(options: InstallOptions): Promise<ProviderId> {
     return initialProvider;
   }
 
-  let selectedProvider: ProviderId;
-  if (options.provider) {
-    selectedProvider = options.provider;
-  } else {
-    const result = await p.select<ProviderId>({
-      message: 'Which LLM provider should claude-mem use to compress observations?',
+  const runClaudeAuthFlow = async (): Promise<void> => {
+    const resolvedAuthMethod = resolveClaudeAuthMethod();
+    const initialAccessMode: ClaudeAccessMode =
+      resolvedAuthMethod === 'subscription' ? 'subscription' : 'api-key';
+
+    const result = await p.select<ClaudeAccessMode>({
+      message: 'Do you use a subscription plan or an API key/gateway for the memory agent?',
       options: [
-        { value: 'claude', label: 'Claude Code auth (default — no extra setup, uses your existing Claude Code subscription)' },
-        { value: 'gemini', label: 'Gemini API key (free tier available — fast and cheap)' },
-        { value: 'openrouter', label: 'OpenRouter API key (BYO model — wide selection of frontier and open models)' },
+        { value: 'subscription', label: 'Subscription plan (recommended — uses your logged-in Claude SDK account)' },
+        { value: 'api-key', label: 'API key or gateway (Anthropic, LiteLLM, or compatible proxy)' },
       ],
-      initialValue: initialProvider,
+      initialValue: initialAccessMode,
     });
 
     if (p.isCancel(result)) {
       p.cancel('Installation cancelled.');
       process.exit(0);
     }
-    selectedProvider = result as ProviderId;
+    if (result === 'subscription') {
+      useSubscriptionAuth();
+      return;
+    }
+
+    const apiModeResult = await p.select<ClaudeApiMode>({
+      message: 'How should claude-mem connect?',
+      options: [
+        { value: 'direct', label: 'Anthropic API key' },
+        { value: 'gateway', label: 'LiteLLM or custom gateway' },
+      ],
+      initialValue: resolvedAuthMethod === 'gateway' || loadClaudeMemEnv().ANTHROPIC_BASE_URL ? 'gateway' : 'direct',
+    });
+
+    if (p.isCancel(apiModeResult)) {
+      p.cancel('Installation cancelled.');
+      process.exit(0);
+    }
+
+    if (apiModeResult === 'gateway') {
+      await configureGateway();
+    } else {
+      await configureDirectApiKey();
+    }
+  };
+
+  let selectedProvider: ProviderId;
+  if (options.provider) {
+    selectedProvider = options.provider;
+  } else {
+    const providerResult = await p.select<ProviderId>({
+      message: 'Which memory provider do you want to use?',
+      options: [
+        { value: 'claude', label: 'Claude Agent SDK (recommended)' },
+        { value: 'gemini', label: 'Gemini' },
+        { value: 'openrouter', label: 'OpenRouter' },
+      ],
+      initialValue: initialProvider,
+    });
+    if (p.isCancel(providerResult)) {
+      p.cancel('Installation cancelled.');
+      process.exit(0);
+    }
+    selectedProvider = providerResult;
   }
 
   if (selectedProvider === 'claude') {
-    persistClaudeProvider();
+    await runClaudeAuthFlow();
     return 'claude';
   }
 
@@ -652,7 +826,7 @@ async function promptProvider(options: InstallOptions): Promise<ProviderId> {
   const apiKeyResult = await p.password({
     message: `Paste your ${providerLabel} API key:`,
     mask: '*',
-    validate: (v: string) => (!v || v.trim().length === 0) ? 'API key required' : undefined,
+    validate: (v?: string) => (!v || v.trim().length === 0) ? 'API key required' : undefined,
   });
 
   if (p.isCancel(apiKeyResult)) {
@@ -678,8 +852,9 @@ async function promptClaudeModel(options: InstallOptions): Promise<void> {
     'claude-sonnet-4-6',
     'claude-opus-4-7',
   ]);
+  const allowCustomModel = resolveClaudeAuthMethod() === 'gateway';
 
-  if (options.model) {
+  if (options.model && !allowCustomModel) {
     if (!allowed.has(options.model)) {
       throw new Error(
         `Unknown Claude model: ${options.model}. Allowed: ${[...allowed].join(', ')}`,
@@ -691,10 +866,39 @@ async function promptClaudeModel(options: InstallOptions): Promise<void> {
     }
     return;
   }
+  if (options.model && allowCustomModel) {
+    const wrote = mergeSettings({ CLAUDE_MEM_MODEL: options.model });
+    if (wrote) {
+      log.info(`Saved gateway model=${options.model} to ~/.claude-mem/settings.json`);
+    }
+    return;
+  }
 
   if (!isInteractive) return;
 
   const initialModel = getSetting('CLAUDE_MEM_MODEL');
+
+  if (allowCustomModel) {
+    const result = await p.text({
+      message: 'Which model should the gateway use?',
+      placeholder: 'claude-haiku-4-5-20251001',
+      defaultValue: initialModel || 'claude-haiku-4-5-20251001',
+      validate: (v?: string) => (!v || v.trim().length === 0) ? 'Model required' : undefined,
+    });
+
+    if (p.isCancel(result)) {
+      p.cancel('Installation cancelled.');
+      process.exit(0);
+    }
+
+    const selectedModel = String(result).trim();
+    const wrote = mergeSettings({ CLAUDE_MEM_MODEL: selectedModel });
+    if (wrote) {
+      log.info(`Saved gateway model=${selectedModel} to ~/.claude-mem/settings.json`);
+    }
+    return;
+  }
+
   const initialValue = allowed.has(initialModel) ? initialModel : 'claude-haiku-4-5-20251001';
 
   const result = await p.select<string>({

src/services/worker/ClaudeProvider.ts

@@ -150,7 +150,7 @@ export class ClaudeProvider {
 
     const settings = SettingsDefaultsManager.loadFromFile(USER_SETTINGS_PATH);
     const maxConcurrent = parseInt(settings.CLAUDE_MEM_MAX_CONCURRENT_AGENTS, 10) || 2;
-    await waitForSlot(maxConcurrent, 60_000);
+    await waitForSlot(maxConcurrent, session.abortController.signal);
 
     const isolatedEnv = sanitizeEnv(await buildIsolatedEnvWithFreshOAuth());
     const authMethod = getAuthMethodDescription();

src/services/worker/http/routes/SessionRoutes.ts

@@ -147,16 +147,16 @@ export class SessionRoutes extends BaseRouteHandler {
 
         const pendingStore = this.sessionManager.getPendingMessageStore();
         try {
-          const cleared = pendingStore.clearPendingForSession(session.sessionDbId);
-          if (cleared > 0) {
-            logger.error('SESSION', `Cleared pending messages after generator error`, {
+          const reset = pendingStore.resetProcessingToPending(session.sessionDbId);
+          if (reset > 0) {
+            logger.warn('SESSION', `Reset processing messages after generator error`, {
               sessionId: session.sessionDbId,
-              cleared
+              reset
             });
           }
         } catch (dbError) {
           const normalizedDbError = dbError instanceof Error ? dbError : new Error(String(dbError));
-          logger.error('HTTP', 'Failed to clear pending messages', {
+          logger.error('HTTP', 'Failed to reset processing messages after generator error', {
             sessionId: session.sessionDbId
           }, normalizedDbError);
         }

src/services/worker/http/routes/SettingsRoutes.ts

@@ -87,6 +87,7 @@ export class SettingsRoutes extends BaseRouteHandler {
       'CLAUDE_MEM_WORKER_PORT',
       'CLAUDE_MEM_WORKER_HOST',
       'CLAUDE_MEM_PROVIDER',
+      'CLAUDE_MEM_CLAUDE_AUTH_METHOD',
       'CLAUDE_MEM_GEMINI_API_KEY',
       'CLAUDE_MEM_GEMINI_MODEL',
       'CLAUDE_MEM_GEMINI_RATE_LIMITING_ENABLED',
@@ -194,6 +195,13 @@ export class SettingsRoutes extends BaseRouteHandler {
       }
     }
 
+    if (settings.CLAUDE_MEM_CLAUDE_AUTH_METHOD) {
+      const validClaudeAuthMethods = ['subscription', 'api-key', 'gateway', 'cli'];
+      if (!validClaudeAuthMethods.includes(settings.CLAUDE_MEM_CLAUDE_AUTH_METHOD)) {
+        return { valid: false, error: 'CLAUDE_MEM_CLAUDE_AUTH_METHOD must be "subscription", "api-key", "gateway", or "cli"' };
+      }
+    }
+
     if (settings.CLAUDE_MEM_GEMINI_MODEL) {
       const validGeminiModels = ['gemini-2.5-flash-lite', 'gemini-2.5-flash', 'gemini-3-flash-preview'];
       if (!validGeminiModels.includes(settings.CLAUDE_MEM_GEMINI_MODEL)) {

src/shared/EnvManager.ts

@@ -13,6 +13,10 @@ export const ENV_FILE_PATH = paths.envFile();
 
 const BLOCKED_ENV_VARS = [
   'ANTHROPIC_API_KEY',       // Issue #733: Prevent auto-discovery from project .env files
+  'ANTHROPIC_AUTH_TOKEN',    // Same leak risk as ANTHROPIC_API_KEY; a token inherited from the
+                             // shell would otherwise short-circuit OAuth lookup at spawn time.
+                             // The fresh token from ~/.claude-mem/.env is re-injected below
+                             // when explicit gateway credentials are configured.
   'CLAUDECODE',              // Prevent "cannot be launched inside another Claude Code session" error
   'CLAUDE_CODE_OAUTH_TOKEN', // Issue #2215: prevent stale parent-process token from leaking into
                              // isolated env. The fresh token is read from the keychain at spawn
@@ -22,6 +26,7 @@ const BLOCKED_ENV_VARS = [
 export interface ClaudeMemEnv {
   ANTHROPIC_API_KEY?: string;
   ANTHROPIC_BASE_URL?: string;
+  ANTHROPIC_AUTH_TOKEN?: string;
   GEMINI_API_KEY?: string;
   OPENROUTER_API_KEY?: string;
 }
@@ -56,7 +61,7 @@ function parseEnvFile(content: string): Record<string, string> {
 function serializeEnvFile(env: Record<string, string>): string {
   const lines: string[] = [
     '# claude-mem credentials',
-    '# This file stores API keys for claude-mem memory agent',
+    '# This file stores keys and gateway settings for the claude-mem memory agent',
     '# Edit this file or use claude-mem settings to configure',
     '',
   ];
@@ -83,6 +88,7 @@ export function loadClaudeMemEnv(): ClaudeMemEnv {
     const result: ClaudeMemEnv = {};
     if (parsed.ANTHROPIC_API_KEY) result.ANTHROPIC_API_KEY = parsed.ANTHROPIC_API_KEY;
     if (parsed.ANTHROPIC_BASE_URL) result.ANTHROPIC_BASE_URL = parsed.ANTHROPIC_BASE_URL;
+    if (parsed.ANTHROPIC_AUTH_TOKEN) result.ANTHROPIC_AUTH_TOKEN = parsed.ANTHROPIC_AUTH_TOKEN;
     if (parsed.GEMINI_API_KEY) result.GEMINI_API_KEY = parsed.GEMINI_API_KEY;
     if (parsed.OPENROUTER_API_KEY) result.OPENROUTER_API_KEY = parsed.OPENROUTER_API_KEY;
 
@@ -126,6 +132,13 @@ export function saveClaudeMemEnv(env: ClaudeMemEnv): void {
       delete updated.ANTHROPIC_BASE_URL;
     }
   }
+  if (env.ANTHROPIC_AUTH_TOKEN !== undefined) {
+    if (env.ANTHROPIC_AUTH_TOKEN) {
+      updated.ANTHROPIC_AUTH_TOKEN = env.ANTHROPIC_AUTH_TOKEN;
+    } else {
+      delete updated.ANTHROPIC_AUTH_TOKEN;
+    }
+  }
   if (env.GEMINI_API_KEY !== undefined) {
     if (env.GEMINI_API_KEY) {
       updated.GEMINI_API_KEY = env.GEMINI_API_KEY;
@@ -171,6 +184,9 @@ export function buildIsolatedEnv(includeCredentials: boolean = true): Record<str
     if (credentials.ANTHROPIC_BASE_URL) {
       isolatedEnv.ANTHROPIC_BASE_URL = credentials.ANTHROPIC_BASE_URL;
     }
+    if (credentials.ANTHROPIC_AUTH_TOKEN) {
+      isolatedEnv.ANTHROPIC_AUTH_TOKEN = credentials.ANTHROPIC_AUTH_TOKEN;
+    }
     if (credentials.GEMINI_API_KEY) {
       isolatedEnv.GEMINI_API_KEY = credentials.GEMINI_API_KEY;
     }
@@ -214,10 +230,15 @@ export async function buildIsolatedEnvWithFreshOAuth(
 
   if (!includeCredentials) return isolatedEnv;
 
-  // If the user already configured an ANTHROPIC_API_KEY in ~/.claude-mem/.env,
-  // honor that and skip OAuth lookup entirely. API key auth is preferred when
-  // explicitly configured because it's stateless and stable.
-  if (isolatedEnv.ANTHROPIC_API_KEY) {
+  // If the user already configured explicit Anthropic/gateway credentials in
+  // ~/.claude-mem/.env, honor those and skip OAuth lookup entirely. A bare
+  // ANTHROPIC_BASE_URL counts because gateways may be tokenless, and falling
+  // back to OAuth would silently route requests to api.anthropic.com.
+  if (
+    isolatedEnv.ANTHROPIC_API_KEY ||
+    isolatedEnv.ANTHROPIC_BASE_URL ||
+    isolatedEnv.ANTHROPIC_AUTH_TOKEN
+  ) {
     clearStaleMarker();
     return isolatedEnv;
   }
@@ -276,10 +297,18 @@ export function hasAnthropicApiKey(): boolean {
   return !!env.ANTHROPIC_API_KEY;
 }
 
+export function hasAnthropicAuthToken(): boolean {
+  const env = loadClaudeMemEnv();
+  return !!env.ANTHROPIC_AUTH_TOKEN;
+}
+
 export function getAuthMethodDescription(): string {
   if (hasAnthropicApiKey()) {
     return 'API key (from ~/.claude-mem/.env)';
   }
+  if (hasAnthropicAuthToken()) {
+    return 'Gateway auth token (from ~/.claude-mem/.env)';
+  }
   // Note: this is a quick sync hint for logging — the authoritative OAuth
   // path is buildIsolatedEnvWithFreshOAuth() which reads the keychain at
   // spawn time. process.env may or may not carry a token here.

src/shared/SettingsDefaultsManager.ts

@@ -74,7 +74,7 @@ export class SettingsDefaultsManager {
     CLAUDE_MEM_WORKER_HOST: '127.0.0.1',
     CLAUDE_MEM_SKIP_TOOLS: 'ListMcpResourcesTool,SlashCommand,Skill,TodoWrite,AskUserQuestion',
     CLAUDE_MEM_PROVIDER: 'claude',  // Default to Claude
-    CLAUDE_MEM_CLAUDE_AUTH_METHOD: 'cli',  // Default to CLI subscription billing (not API key)
+    CLAUDE_MEM_CLAUDE_AUTH_METHOD: 'subscription',  // Default to logged-in Claude SDK auth (not API key)
     CLAUDE_MEM_GEMINI_API_KEY: '',  // Empty by default, can be set via UI or env
     CLAUDE_MEM_GEMINI_MODEL: 'gemini-2.5-flash-lite',  // Default Gemini model (highest free tier RPM)
     CLAUDE_MEM_GEMINI_RATE_LIMITING_ENABLED: 'true',  // Rate limiting ON by default for free tier users

src/supervisor/process-registry.ts

@@ -174,9 +174,11 @@ export class ProcessRegistry {
 
   unregister(id: string): void {
     this.initialize();
+    const existing = this.entries.get(id);
     this.entries.delete(id);
     this.runtimeProcesses.delete(id);
     this.persist();
+    if (existing?.type === 'sdk') notifySlotAvailable();
   }
 
   clear(): void {
@@ -213,16 +215,19 @@ export class ProcessRegistry {
     this.initialize();
 
     let removed = 0;
+    let removedSdk = 0;
     for (const [id, info] of this.entries) {
       if (isPidAlive(info.pid)) continue;
       this.entries.delete(id);
       this.runtimeProcesses.delete(id);
       removed += 1;
+      if (info.type === 'sdk') removedSdk += 1;
     }
 
     if (removed > 0) {
       this.persist();
     }
+    for (let i = 0; i < removedSdk; i += 1) notifySlotAvailable();
 
     return removed;
   }
@@ -321,6 +326,9 @@ export class ProcessRegistry {
       this.runtimeProcesses.delete(record.id);
     }
     this.persist();
+    for (const record of sessionRecords) {
+      if (record.type === 'sdk') notifySlotAvailable();
+    }
 
     logger.info('SYSTEM', `Reaped ${sessionRecords.length} process(es) for session ${sessionId}`, {
       sessionId: sessionIdNum,
@@ -428,6 +436,7 @@ export async function ensureSdkProcessExit(
 }
 
 const TOTAL_PROCESS_HARD_CAP = 10;
+const SLOT_RECHECK_INTERVAL_MS = 5_000;
 const slotWaiters: Array<() => void> = [];
 
 function getActiveSdkCount(): number {
@@ -439,7 +448,8 @@ function notifySlotAvailable(): void {
   if (waiter) waiter();
 }
 
-export async function waitForSlot(maxConcurrent: number, timeoutMs: number = 60_000): Promise<void> {
+export async function waitForSlot(maxConcurrent: number, signal?: AbortSignal): Promise<void> {
+  getProcessRegistry().pruneDeadEntries();
   const activeCount = getActiveSdkCount();
   if (activeCount >= TOTAL_PROCESS_HARD_CAP) {
     throw new Error(`Hard cap exceeded: ${activeCount} processes in registry (cap=${TOTAL_PROCESS_HARD_CAP}). Refusing to spawn more.`);
@@ -447,25 +457,55 @@ export async function waitForSlot(maxConcurrent: number, timeoutMs: number = 60_
 
   if (activeCount < maxConcurrent) return;
 
+  if (signal?.aborted) {
+    throw new Error('waitForSlot aborted before queuing');
+  }
+
   logger.info('PROCESS', `Pool limit reached (${activeCount}/${maxConcurrent}), waiting for slot...`);
 
   return new Promise<void>((resolve, reject) => {
-    const timeout = setTimeout(() => {
+    let recheckTimer: ReturnType<typeof setInterval> | null = null;
+    let abortHandler: (() => void) | null = null;
+    const cleanup = () => {
+      if (recheckTimer) clearInterval(recheckTimer);
+      if (abortHandler && signal) signal.removeEventListener('abort', abortHandler);
       const idx = slotWaiters.indexOf(onSlot);
       if (idx >= 0) slotWaiters.splice(idx, 1);
-      reject(new Error(`Timed out waiting for agent pool slot after ${timeoutMs}ms`));
-    }, timeoutMs);
-
+    };
     const onSlot = () => {
-      clearTimeout(timeout);
-      if (getActiveSdkCount() < maxConcurrent) {
+      const count = getActiveSdkCount();
+      if (count >= TOTAL_PROCESS_HARD_CAP) {
+        cleanup();
+        reject(new Error(`Hard cap exceeded: ${count} processes in registry (cap=${TOTAL_PROCESS_HARD_CAP}). Refusing to spawn more.`));
+        return;
+      }
+
+      if (count < maxConcurrent) {
+        cleanup();
         resolve();
       } else {
         slotWaiters.push(onSlot);
       }
     };
 
+    if (signal) {
+      abortHandler = () => {
+        cleanup();
+        reject(new Error('waitForSlot aborted'));
+      };
+      signal.addEventListener('abort', abortHandler, { once: true });
+    }
+
     slotWaiters.push(onSlot);
+    recheckTimer = setInterval(() => {
+      const removed = getProcessRegistry().pruneDeadEntries();
+      if (removed > 0) {
+        logger.info('PROCESS', 'Pruned stale process registry entries while waiting for agent slot', { removed });
+        return;
+      }
+      notifySlotAvailable();
+    }, SLOT_RECHECK_INTERVAL_MS);
+    recheckTimer.unref?.();
   });
 }
 
@@ -565,7 +605,6 @@ export function spawnSdkProcess(
       logger.warn('SDK_SPAWN', `[session-${sessionDbId}] Claude process exited`, { code, signal, pid });
     }
     registry.unregister(recordId);
-    notifySlotAvailable();
   });
 
   if (!child.stdin || !child.stdout || !child.stderr) {