agentleFS
Sign inStart free
arndvsarndvs/ctrlshft/skills/explore/SKILL.md

explore

Skill0 starsChanged 6 months agoMITView on GitHub

Deep codebase exploration using parallel subagents. Use when asked to 'explore', 'understand', 'investigate', 'map out', 'how does X work', or 'audit' a part of the codebase.

We split this skill into the rules it gives an agent

5 rules. 0 are common, 3 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 55 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If running interactively (human present), output "Read Explore skill." to acknowledge.

    line 8
  2. Very specific:

    Always use the search_subagent tool and the word explore when spawning sub-agents — this triggers deep file traversal mode to run MULTIPLE …

    line 43
  3. Distinctive:

    Do not rely on a single agent to cover the full topic

    line 45
  4. Distinctive:

    After all sub-agents complete, produce a unified summary with cross-references between areas

    line 46
  5. Distinctive:

    If a sub-agent surfaces something unexpected, spawn an additional agent to follow up

    line 47

“Common” is measured: kept by at least 6 in 10 of the 55 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

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

  1. HUD event — emit a start event so the HUD tracks this exploration:
    source ~/dotfiles/bin/write-hud-state.sh
write_hud_event "info" "explore: started — $TOPIC"
    Emit again at synthesis (explore: completed — N sub-agents, M files covered).
  2. Decompose the topic into distinct areas of concern
  3. Spawn a dedicated sub-agent for each area using the explore keyword in the sub-agent prompt
  4. Synthesize the results from all sub-agents into a single cohesive summary

Example decomposition for "How does authentication work?":

  • Explore how sessions are created and stored
  • Explore how middleware protects routes
  • Explore how tokens are issued and validated
  • Explore how logout and expiry are handled

Rules

  • Always use the search_subagent tool and the word explore when 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.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free