kimi:explore

/kimi:explore

Structured codebase analysis: Executive Summary + Health Score + Architecture Deep Dive.

Usage

/kimi:explore how does the auth module work?
/kimi:explore find all database connection code
/kimi:explore --background give me an architecture overview
/kimi:explore                     # generic analysis when no question is given

Process

1. Load explore prompt template

   Read("plugins/kimi/prompts/explore.md")
   

2. Dispatch to Kimi

   • The free-text question is the primary input; when no question is given, fall back to a generic codebase analysis.
   • Use the explore role (--role explore).

   Bash("node plugins/kimi/scripts/broker.mjs dispatch \
     --prompt '<question>' \
     --role explore \
     --mode explore \
     [--background]")
   

   Fallback prompt when no question: "Analyze the codebase at $(pwd). Follow the explore prompt template."

3. Parse and render

   • Parse JSON output for summary, healthScore, techStack, insights, architecture.
   • Render as a markdown report.

Output

## 🎯 Executive Summary
...
### Health Score: 8/10
...
### Tech Stack
...
### Key Insights
...

Notes

• Runs under the broker's --role explore: the read-only role prompt in plugins/kimi/roles/explore.md is composed into the head of the -p prompt (--agent-file YAMLs no longer exist in Kimi Code 0.x).
• Read-only enforcement is a prompt-level contract — 0.x cannot exclude tools headlessly.