Codebase exploration and brainstorming in interactive Q&A format. Delegates research to a sub-agent, accumulates findings in a summary chain, and saves an exploration log when done.
$ARGUMENTS — the first question or topic. If empty, asks via AskUserQuestion.
/sp:explore how does authorization work
/sp:explore compare caching approaches in the project
/sp:explore what if we replaced REST with gRPC
| Phase | Name | What happens |
|---|---|---|
| 1 | Init | Pick slug, initialize summary chain and QA log |
| 2 | Loop | User-driven Q&A: enrich the prompt with context, dispatch explore-agent |
| 3 | Finalize | Sub-agent writes the exploration log |
| 4 | Complete | AskUserQuestion: another question / create a task via /sp:task / finish |
Exploration log: docs/ai/<slug>/<slug>-exploration.md.
Each Q&A entry contains structured sections: question context, answer, details (file:line, code snippets), alternatives (brainstorm), key files, sources. A final summary closes the log.
| Agent | Model | Role |
|---|---|---|
explore-agent |
sonnet | Exploration: code search, analysis, structured-output answer |
explore-log-writer |
haiku | Writes a structured exploration log to docs/ai/<slug>/<slug>-exploration.md |
/sp:explore how does the notification system work
Result: a Q&A session that accumulates context; on completion, an exploration log.
Independent skill. After exploration, use /sp:task to create a task from the findings.