fix: comprehensive tool argument normalization hardening

- Remove all { raw: ... } returns that caused InputValidationError with
  z.strictObject schemas — return {} instead for clean Zod errors
- Extend normalizeAtStop buffering to all mapped tools (Read, Write,
  Edit, Glob, Grep) so streaming paths also get normalized
- Make repairPossiblyTruncatedObjectJson generic — repair any valid
  JSON object, not just ones with a command field
- Export hasToolFieldMapping for streaming normalizeAtStop decision
- Skip normalization on finish_reason: length to preserve raw truncated
  buffer
- Update all test expectations to match new behavior

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

src/services/api/openaiShim.test.ts

@@ -744,7 +744,7 @@ test('keeps terminal empty Bash tool arguments invalid in non-streaming response
       type: 'tool_use',
       id: 'function-call-1',
       name: 'Bash',
-      input: { raw: '' },
+      input: {},
     },
   ])
 })
@@ -1094,7 +1094,7 @@ test('keeps terminal whitespace-only Bash arguments invalid in streaming respons
     .map(event => (event.delta as Record<string, unknown>).partial_json)
     .join('')
 
-  expect(normalizedInput).toBe('{"raw":" "}')
+  expect(normalizedInput).toBe('{}')
 })
 
 test('normalizes streaming Bash arguments that begin with bracket syntax', async () => {
@@ -1423,7 +1423,7 @@ test('does not normalize incomplete streamed Bash commands when finish_reason is
   expect(streamedInput).toBe('rg --fi')
 })
 
-test('does not repair truncated Bash objects that do not contain command', async () => {
+test('repairs truncated JSON objects even without command field', async () => {
   globalThis.fetch = (async (_input, _init) => {
     const chunks = makeStreamChunks([
       {
@@ -1496,7 +1496,7 @@ test('does not repair truncated Bash objects that do not contain command', async
     .map(event => (event.delta as Record<string, unknown>).partial_json)
     .join('')
 
-  expect(streamedInput).toBe('{"raw":"{\\"cwd\\":\\"/tmp\\""}')
+  expect(streamedInput).toBe('{"cwd":"/tmp"}')
 })
 
 test('preserves raw input for unknown plain string tool arguments', async () => {
@@ -1554,7 +1554,7 @@ test('preserves raw input for unknown plain string tool arguments', async () =>
       type: 'tool_use',
       id: 'function-call-1',
       name: 'UnknownTool',
-      input: { raw: 'pwd' },
+      input: {},
     },
   ])
 })

src/services/api/openaiShim.ts

@@ -44,6 +44,7 @@ import { sanitizeSchemaForOpenAICompat } from '../../utils/schemaSanitizer.js'
 import { redactSecretValueForDisplay } from '../../utils/providerProfile.js'
 import {
   normalizeToolArguments,
+  hasToolFieldMapping,
 } from './toolArgumentNormalization.js'
 
 type SecretValueSource = Partial<{
@@ -486,7 +487,7 @@ const JSON_REPAIR_SUFFIXES = [
 function repairPossiblyTruncatedObjectJson(raw: string): string | null {
   try {
     const parsed = JSON.parse(raw)
-    return parsed && typeof parsed === 'object' && !Array.isArray(parsed) && typeof (parsed as Record<string, unknown>).command === 'string'
+    return parsed && typeof parsed === 'object' && !Array.isArray(parsed)
       ? raw
       : null
   } catch {
@@ -494,12 +495,7 @@ function repairPossiblyTruncatedObjectJson(raw: string): string | null {
       try {
         const repaired = raw + combo
         const parsed = JSON.parse(repaired)
-        if (
-          parsed &&
-          typeof parsed === 'object' &&
-          !Array.isArray(parsed) &&
-          typeof (parsed as Record<string, unknown>).command === 'string'
-        ) {
+        if (parsed && typeof parsed === 'object' && !Array.isArray(parsed)) {
           return repaired
         }
       } catch {}
@@ -619,7 +615,7 @@ async function* openaiStreamToAnthropic(
 
               const toolBlockIndex = contentBlockIndex
               const initialArguments = tc.function.arguments ?? ''
-              const normalizeAtStop = tc.function.name === 'Bash'
+              const normalizeAtStop = hasToolFieldMapping(tc.function.name)
               activeToolCalls.set(tc.index, {
                 id: tc.id,
                 name: tc.function.name,

src/services/api/toolArgumentNormalization.test.ts

@@ -19,20 +19,18 @@ describe('normalizeToolArguments', () => {
       ).toEqual({ command: 'echo hi' })
     })
 
-    test('returns { raw } for blank string', () => {
-      expect(normalizeToolArguments('Bash', '')).toEqual({ raw: '' })
-      expect(normalizeToolArguments('Bash', '   ')).toEqual({ raw: '   ' })
+    test('returns empty object for blank string', () => {
+      expect(normalizeToolArguments('Bash', '')).toEqual({})
+      expect(normalizeToolArguments('Bash', '   ')).toEqual({})
     })
 
-    test('returns { raw } for JSON-encoded blank string', () => {
-      expect(normalizeToolArguments('Bash', '""')).toEqual({ raw: '' })
-      expect(normalizeToolArguments('Bash', '"  "')).toEqual({ raw: '  ' })
+    test('returns parsed blank for JSON-encoded blank string', () => {
+      expect(normalizeToolArguments('Bash', '""')).toEqual('')
+      expect(normalizeToolArguments('Bash', '"  "')).toEqual('  ')
     })
 
-    test('returns { raw } for likely structured object literal that fails parse', () => {
-      expect(normalizeToolArguments('Bash', '{ "command": "pwd"')).toEqual({
-        raw: '{ "command": "pwd"',
-      })
+    test('returns empty object for malformed structured object literal', () => {
+      expect(normalizeToolArguments('Bash', '{ "command": "pwd"')).toEqual({})
     })
 
     test.each([
@@ -40,9 +38,9 @@ describe('normalizeToolArguments', () => {
       ["{'command':'pwd'}"],
       ['{command: pwd}'],
     ])(
-      'returns { raw } for malformed object-shaped string %s (does not wrap into command)',
+      'returns empty object for malformed object-shaped string %s (does not wrap into command)',
       (input) => {
-        expect(normalizeToolArguments('Bash', input)).toEqual({ raw: input })
+        expect(normalizeToolArguments('Bash', input)).toEqual({})
       },
     )
 
@@ -157,10 +155,8 @@ describe('normalizeToolArguments', () => {
   })
 
   describe('unknown tools', () => {
-    test('returns { raw } for plain string', () => {
-      expect(normalizeToolArguments('UnknownTool', 'some value')).toEqual({
-        raw: 'some value',
-      })
+    test('returns empty object for plain string (no known field mapping)', () => {
+      expect(normalizeToolArguments('UnknownTool', 'some value')).toEqual({})
     })
 
     test('passes through structured JSON object', () => {
@@ -175,7 +171,7 @@ describe('normalizeToolArguments', () => {
       expect(normalizeToolArguments('UnknownTool', '[]')).toEqual([])
     })
 
-    test('returns JSON-encoded string as parsed string for unknown tools', () => {
+    test('returns parsed string for JSON-encoded string on unknown tools', () => {
       expect(normalizeToolArguments('UnknownTool', '"hello"')).toEqual(
         'hello',
       )

src/services/api/toolArgumentNormalization.ts

@@ -26,6 +26,10 @@ function getPlainStringToolArgumentField(toolName: string): string | null {
   return STRING_ARGUMENT_TOOL_FIELDS[toolName] ?? null
 }
 
+export function hasToolFieldMapping(toolName: string): boolean {
+  return toolName in STRING_ARGUMENT_TOOL_FIELDS
+}
+
 function wrapPlainStringToolArguments(
   toolName: string,
   value: string,
@@ -46,25 +50,20 @@ export function normalizeToolArguments(
     if (isRecord(parsed)) {
       return parsed
     }
-    if (toolName === 'Bash') {
-      if (typeof parsed === 'string') {
-        if (isBlankString(parsed)) {
-          return { raw: parsed }
-        }
-        return wrapPlainStringToolArguments(toolName, parsed) ?? parsed
-      }
-      return parsed
-    }
-    if (typeof parsed === 'string') {
+    // Parsed as a non-object JSON value (string, number, boolean, null, array)
+    if (typeof parsed === 'string' && !isBlankString(parsed)) {
       return wrapPlainStringToolArguments(toolName, parsed) ?? parsed
     }
+    // For blank strings, booleans, null, arrays — pass through as-is
+    // and let Zod schema validation produce a meaningful error
     return parsed
   } catch {
+    // rawArguments is not valid JSON — treat as a plain string
     if (isBlankString(rawArguments) || isLikelyStructuredObjectLiteral(rawArguments)) {
-      return { raw: rawArguments }
+      // Blank or looks like a malformed object literal — don't wrap into
+      // a tool field to avoid turning garbage into executable input
+      return {}
     }
-    return (
-      wrapPlainStringToolArguments(toolName, rawArguments) ?? { raw: rawArguments }
-    )
+    return wrapPlainStringToolArguments(toolName, rawArguments) ?? {}
   }
 }