No actionable comments were generated in the recent review. 🎉

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

_{Comment @coderabbitai help to get the list of available commands.}

Thanks for adding this. I think the feature is useful, but I want to clarify the behavior before we merge because prompts are already exported today in trace data.

Currently prompt text is attached to OTel spans via OpenInference attributes:

• src/handlers/session.ts sets input.value and llm.input_messages on run/session spans
• src/handlers/message.ts sets input.value and llm.input_messages on LLM spans

So this PR does not add prompt capture for the first time. It adds an opt-in copy of the prompt into the user_prompt OTLP log event.

That distinction matters for privacy/security expectations. With this change, OPENCODE_CAPTURE_PROMPT only controls whether prompts appear in logs. It does not prevent prompt text from being exported in traces when traces are enabled.

Could you update the README/startup wording to make that explicit? Something like:

OPENCODE_CAPTURE_PROMPT controls whether full prompt text is also included in user_prompt log events. Prompt text may already be exported as trace span attributes unless traces are disabled with OPENCODE_DISABLE_TRACES.

Also, repo convention is that config values are available both as OPENCODE_* env vars and plugin tuple options with option → env → default precedence. Could you add a matching capturePrompt?: boolean option to OtelPluginOptions, wire it through loadConfig, document it in the plugin options table, and add a small precedence test?

I’m supportive of the feature as log enrichment, but I’d like the naming/docs to avoid implying this is the only control for prompt export.