ワンクリックで
rounds-investigate
Investigate a distributed trace by ID — fetches the full trace, reads source files, and explains the end-to-end code flow
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Investigate a distributed trace by ID — fetches the full trace, reads source files, and explains the end-to-end code flow
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Assess telemetry data — identifies distinct transaction types from an explore query and launches sub-agents to analyze each for instrumentation quality, code efficiency, usage correctness, and skill improvement opportunities
Print available rounds skills, project directory layout, and a brief overview of what rounds does
Exploratory querying of telemetry data — search logs by keyword/metadata, search spans by metadata, and build full transaction trees from trace IDs using the rounds adapter interfaces
Get full details for a single rounds error signature including diagnosis, recent events, and related signatures
List error signatures in the rounds database, optionally filtered by status
Mute a rounds error signature to suppress further notifications
| name | rounds-investigate |
| description | Investigate a distributed trace by ID — fetches the full trace, reads source files, and explains the end-to-end code flow |
| user_invocable | true |
| args | ["trace_id"] |
Investigate a distributed trace using rounds' trace investigation pipeline. Fetches the full trace from the telemetry backend, reads the relevant source files from the mounted codebase, and explains the end-to-end code flow with actual file:line citations.
/rounds-investigate TRACE_ID
TRACE_ID is a 32-char lowercase hex OpenTelemetry trace ID, e.g.:
d5b99e396c3aeac81aa6074635254687
Run the following, substituting $ARGUMENTS for the trace ID:
cd /workspace/rounds && python -m rounds.main cli-run investigate-trace $ARGUMENTS
The command outputs a JSON object. Parse it and present the results as follows:
On success ("status": "success"):
investigation.summary field as a clear paragraphinvestigation.code_flow as a numbered list; each entry is a
step in the request pathinvestigation.services_involved as a comma-separated
listinvestigation.key_findings as a bulleted listinvestigation.cost_usd formatted as $X.XXXXOn error ("status": "error"):
Show the message field and suggest:
CODEBASE_PATH=/workspace/target is set and the target is mounted/rounds-investigate d5b99e396c3aeac81aa6074635254687