### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? Prompt-type hooks (`"type": "prompt"`) fail on Vertex AI. Claude Code sends an `output_config` parameter that the Vertex AI API does not support. Environ

[BUG] Prompt-type hooks fail on Vertex AI with "output_config: Extra inputs are not permitted"

Status Fixed / completed

Reported on v2.1.81

Maintainer reply ✓ Yes — mhegazy

Workaround ✓ Mentioned in thread ↓

Activity 5 comments · opened Mar 23, 2026 · closed Apr 27, 2026

💡 Likely answer: A maintainer (mhegazy, contributor) responded on this thread — see the highlighted reply below.

Preflight Checklist

[x] I have searched existing issues and this hasn't been reported yet
[x] This is a single bug report (please file separate reports for different bugs)
[x] I am using the latest version of Claude Code

What's Wrong?

Prompt-type hooks ("type": "prompt") fail on Vertex AI. Claude Code sends an output_config parameter that the Vertex AI API does not support.

Environment: Claude Code v2.1.81, Vertex AI

What Should Happen?

The prompt-based hook should run successfully on Vertex AI API.

Error Messages/Logs

[DEBUG] Hooks: Processing prompt hook with prompt: Summarize what this command did: {"session_id":"...","tool_name":"Bash","tool_input":{"command":"ls ..."},...}
[DEBUG] Hooks: Querying model with 1 messages
[ERROR] API error (attempt 1/11): 400 {"type":"error","error":{"type":"invalid_request_error","message":"output_config: Extra inputs are not permitted"}}
[DEBUG] Hooks: Model response: API Error: 400 ...
[DEBUG] Hooks: error parsing response as JSON: API Error: 400 ...

Steps to Reproduce

Configure Claude Code to use Vertex AI (CLAUDE_CODE_USE_VERTEX=1)
Add a prompt-type hook to .claude/settings.json:

  {
    "hooks": {
      "PostToolUse": [
        {
          "matcher": "Bash",
          "hooks": [
            {
              "type": "prompt",
              "prompt": "Summarize what this command did: $ARGUMENTS",
              "statusMessage": "Analyzing..."
            }
          ]
        }
      ]
    }
  }

Run Claude Code with --debug
Ask Claude to run any command (e.g., ls)
Observe the hook error in the debug log

Claude Model

Other

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.81 (Claude Code)

Platform

Google Vertex AI

Operating System

macOS

Terminal/Shell

Other

Additional Information

_No response_

View original on GitHub ↗

5 Comments

yurukusa · 6 months ago

This is the same output_config issue as #41149. Vertex AI's streamRawPredict endpoint rejects the parameter.

// Instead of:
{"type": "prompt", "prompt": "Summarize what this command did: ..."}
// Use:
{"type": "command", "command": "echo 'Hook triggered' >&2"}

Command hooks (type: command) run as shell scripts and don't make API calls, so they work regardless of backend. Prompt hooks (type: prompt) make a separate API call to the model, which is where the output_config incompatibility hits.
If you specifically need model-powered analysis in your hooks, you can call the API directly from a command hook:

curl -s https://your-vertex-endpoint/... -d '{"messages": ...}'

jeremyeder · 5 months ago

Still a bug in 2.1.114. Changing to command type happens to be OK in my current scenario.

jarle73STB · 5 months ago

Still hitting this on Vertex AI. Converting to command hooks isn't a realistic workaround for us — we manage hook config centrally across 40+ repos, and prompt hooks work great for that since it's just JSON.
Command hooks would mean distributing and maintaining scripts in every repo. Would appreciate a proper fix for this.

mhegazy contributor · 5 months ago

Should be fixed in next release

github-actions[bot] · 5 months ago

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.