arndvs Explore
If running interactively (human present), output "Read Explore skill." to acknowledge. If running with --dangerously-skip-permissions (AFK/unattended), skip acknowledgement and proceed directly.
When tasked with understanding, auditing, or investigating any part of a codebase, spawn multiple sub-agents using the explore verb to maximize file coverage and depth. Use the search_subagent tool with explore in the prompt to trigger deep traversal mode, allowing the agent to recursively search through directories and files for relevant information.
When to Use This
Trigger this approach when asked to:
- Understand how a feature works
- Audit or review a codebase
- Investigate a bug or behavior
- Map out data flow or architecture
- Prepare before building a new feature
How to Execute
- HUD event — emit a start event so the HUD tracks this exploration:
Emit again at synthesis (source ~/dotfiles/bin/write-hud-state.sh write_hud_event "info" "explore: started — $TOPIC"explore: completed — N sub-agents, M files covered). - Decompose the topic into distinct areas of concern
- Spawn a dedicated sub-agent for each area using the
explorekeyword in the sub-agent prompt - Synthesize the results from all sub-agents into a single cohesive summary
Example decomposition for "How does authentication work?":
Explorehow sessions are created and storedExplorehow middleware protects routesExplorehow tokens are issued and validatedExplorehow logout and expiry are handled
Rules
- Always use the
search_subagenttool and the wordexplorewhen spawning sub-agents — this triggers deep file traversal mode to run MULTIPLE sub-agents in PARALLEL, maximizing coverage and depth - Each sub-agent should have a narrow, specific focus — avoid broad prompts per agent
- Do not rely on a single agent to cover the full topic
- After all sub-agents complete, produce a unified summary with cross-references between areas
- If a sub-agent surfaces something unexpected, spawn an additional agent to follow up
Why This Matters
A single agent pass typically reads 6–10 files. A dedicated explore sub-agent will aggressively search the repo, calling 20–30+ tools and covering far more ground. For any non-trivial exploration task, a single pass will produce an incomplete picture.