What this error means
Stage 1 failed to produce valid 'concerns' array after 3 attempts — aborting review is a Vertex AI / Claude failure pattern reported for developers trying to fix vertex ai claude provider silently dropping response_format field, causing structured json output failures. Based on the imported evidence, treat this as a tool-specific troubleshooting page rather than a generic API error.
Why this happens
Sashiko issue #171: Vertex AI Claude provider (claude-sonnet-4-6 and claude-opus-4-6 in us-east5) drops response_format field in generate_claude(), causing Stage 1 AI review to always fail with 'failed to produce valid concerns array after 3 attempts'. Direct rawPredict calls with same credentials work fine — the bug is in the SDK layer not forwarding response_format to Vertex ClaudeRequest. 12 comments.
Common causes
- When using Claude models through Google Vertex AI (not direct Anthropic API), the response_format field is silently dropped from requests. This causes any tool or application that expects structured JSON output (via response_format) to fail with parsing errors. The Vertex API endpoint itself works fine for free-form responses, making this a subtle integration bug that only surfaces with structured output requirements.
- Sashiko issue #171: Vertex AI Claude provider (claude-sonnet-4-6 and claude-opus-4-6 in us-east5) drops response_format field in generate_claude(), causing Stage 1 AI review to always fail with 'failed to produce valid concerns array after 3 attempts'. Direct rawPredict calls with same credentials work fine — the bug is in the SDK layer not forwarding response_format to Vertex ClaudeRequest. 12 comments.
Quick fixes
- Confirm the exact error signature matches
Stage 1 failed to produce valid 'concerns' array after 3 attempts — aborting review. - Check the Vertex AI / Claude account, local tool state, and provider configuration involved in the failing workflow.
- Compare the failing environment with a known working setup, then change one configuration value at a time.
Platform/tool-specific checks
- Verify the command, editor, extension, or API client that produced the error.
- Compare local settings with CI, deployment, or editor-level settings when the error appears in only one environment.
- Avoid deleting credentials, local model data, or project settings until the failing scope is clear.
Step-by-step troubleshooting
- Capture the exact error message and the command, editor action, or request that triggered it.
- Check whether the failure is account/auth, quota/rate, model/provider, local runtime, or deployment configuration.
- Review the source evidence below and compare it with your environment.
- Apply one change at a time and rerun the smallest failing action.
- Keep the working fix documented for the team or deployment environment.
How to prevent it
- Keep provider/tool configuration documented.
- Record non-secret diagnostics such as tool version, provider name, model name, and command path.
- Add a lightweight check before CI or production workflows depend on the tool.