feat: add auto-fix service — auto-lint and test after AI file edits

[gnanam1990]

Summary

Adds a built-in auto-fix service that automatically runs user-configured lint and test commands after AI file edits (FileEdit, FileWrite). When errors are detected, they are fed back to the AI as additional context so it can self-repair — closing the manual feedback loop that currently requires users to run lint/test, copy errors, and paste them back.

Inspired by Aider's auto-lint-and-test loop, which is their #1 user satisfaction driver.

What's included

• Config schema (src/services/autoFix/autoFixConfig.ts) — Zod-validated settings with enabled, lint, test, maxRetries (0–10, default 3), and timeout (1–300s, default 30s). Rejects enabled configs with no lint or test command.

• Runner (src/services/autoFix/autoFixRunner.ts) — Executes lint/test commands via child_process.spawn, captures stdout/stderr, enforces timeouts, skips tests if lint fails first, and builds an AI-readable error summary.

• Hook integration (src/services/autoFix/autoFixHook.ts) — shouldRunAutoFix() gates execution to file_edit/file_write tools only. buildAutoFixContext() wraps errors in <auto_fix_feedback> tags that the AI can act on.

• PostToolUse wiring (src/services/tools/toolHooks.ts) — After the existing PostToolUse hook loop, runs the auto-fix check and yields errors as hook_additional_context attachment messages.

• /auto-fix slash command (src/commands/auto-fix.ts) — Guided setup wizard that helps users configure their lint/test commands in settings.

• Settings schema (src/utils/settings/types.ts) — Added optional autoFix field to SettingsSchema.

Configuration

Add to .claude/settings.json or .claude/settings.local.json:

{
  "autoFix": {
    "enabled": true,
    "lint": "eslint . --fix",
    "test": "bun test",
    "maxRetries": 3,
    "timeout": 30000
  }
}

Or run /auto-fix for guided setup.

How it works

AI edits file → lint runs → errors? → feed back to AI → AI fixes → re-lint
                         ↓ no errors
                    tests run → failures? → feed back to AI → AI fixes → re-test
                             ↓ no failures
                        clean result delivered

Test coverage

• 29 tests across 4 test files (config, runner, hook, integration)
• 541/541 full suite tests pass
• Zero new type errors
• Build succeeds

Files changed

File	Change
src/services/autoFix/autoFixConfig.ts	New — config schema and reader
src/services/autoFix/autoFixConfig.test.ts	New — 11 tests
src/services/autoFix/autoFixRunner.ts	New — command execution engine
src/services/autoFix/autoFixRunner.test.ts	New — 8 tests
src/services/autoFix/autoFixHook.ts	New — hook helpers
src/services/autoFix/autoFixHook.test.ts	New — 8 tests
src/services/autoFix/autoFixIntegration.test.ts	New — 2 e2e tests
src/commands/auto-fix.ts	New — slash command
src/commands.ts	Register /auto-fix command
src/services/tools/toolHooks.ts	Wire auto-fix into PostToolUse
src/utils/settings/types.ts	Add autoFix to settings schema