Skip to content

/exploration

Use when exploring an unfamiliar codebase -- architecture analysis, pattern detection, dependency mapping, rapid discovery.

shell
$ npx -y skills add fusengine/agents --skill exploration --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.
  • You can call itInvoke it directly when you want it.
  • Slash command/exploration
How auto-invocation works

Context preview

The summary Claude sees to decide when to auto-load this skill.

Use when exploring an unfamiliar codebase -- architecture analysis, pattern detection, dependency mapping, rapid discovery.

SKILL.md

exploration.SKILL.md
name: exploration
description: Use when exploring an unfamiliar codebase -- architecture analysis, pattern detection, dependency mapping, rapid discovery.
argument-hint: "[directory] [--thorough]"
context: fork
agent: fuse-ai-pilot:explore-codebase
user-invocable: false

<objective> Exploration runs a 5-phase reconnaissance protocol for rapid codebase discovery: initial reconnaissance (root files, config, entry points), structure mapping, entry-point detection per language, dependency analysis (package.json/pyproject.toml/go.mod/Cargo.toml/composer.json), and architecture-pattern detection (MVC, Clean, Hexagonal, feature-based, layered, modular monolith, Next.js App Router, and more).

It is evidence-based only -- no assumptions without code proof -- and produces a structured report of tech stack, entry points, and detected architecture pattern for whoever consumes it (typically the APEX Analyze phase or a domain expert). </objective>

**Session:** ${CLAUDE_SESSION_ID}

Exploration Skill

Exploration Protocol

Run these 5 phases in order — full commands in `references/exploration-protocol.md`:

1. **Initial Reconnaissance** — list root files, find config files, check common entry points 2. **Structure Mapping** — tree view (excluding noise dirs), identify main directories 3. **Entry Points Detection** — grep for `main`/`index`/server bootstrap per language 4. **Dependency Analysis** — read `package.json`/`pyproject.toml`/`go.mod`/`Cargo.toml`/`composer.json` 5. **Pattern Detection** — check for MVC, Clean/Hexagonal, feature-based, Next.js App Router directories

---

Architecture Pattern Detection

Pattern Indicators

| Pattern | Key Directories | Indicators | |---------|-----------------|------------| | **MVC** | `controllers/`, `models/`, `views/` | Rails, Laravel, Express | | **Clean Architecture** | `domain/`, `application/`, `infrastructure/` | DDD, Use cases | | **Hexagonal** | `adapters/`, `ports/`, `core/` | Ports & adapters | | **Feature-based** | `features/[name]/` | All layers per feature | | **Layered** | `presentation/`, `business/`, `data/` | Traditional 3-tier | | **Monolith** | Single `src/` | Mixed concerns | | **Microservices** | Multiple `services/` | Separate repos/folders | | **Next.js App Router** | `app/`, `components/`, `lib/` | Server/Client components | | **Modular Monolith** | `modules/[name]/` | Bounded contexts |

---

Forbidden Behaviors

  • ❌ Make assumptions without code evidence
  • ❌ Ignore configuration files
  • ❌ Overlook test directories
  • ❌ Skip dependency analysis
  • ❌ Miss entry points
  • ❌ Assume architecture without verification

Behavioral Traits

  • Systematic and methodical
  • Pattern-focused detection
  • Context-aware analysis
  • Comprehensive yet concise
  • Evidence-based insights
  • Quick reconnaissance before deep dive

---

Detailed References (Load on Demand)

  • `references/exploration-protocol.md` — Load when running the 5-phase protocol (exact shell commands per phase)
  • `references/tech-stack-detection.md` — Load when identifying framework/ORM/tooling per language (JS/TS, Python, Go, PHP, Rust)
  • `references/code-organization-checks.md` — Load when auditing interface separation, business-logic location, or state management
  • `references/response-format.md` — Load when writing the final exploration report
  • `references/quick-analysis-commands.md` — Load for one-liner full-stack assessment, file counts, or large-file detection
Read more
Read it on GitHub ↗
Ships withfusengine-agents

A plugin ecosystem that turns Claude Code into a supervised, multi-agent development environment.

Get the whole plugin, auto-invoked
Stats
22
Stars
0
Views
3
Forks
Active
Maintenance
CSS
Language
MIT
License
1d ago
Last commit
6mo ago
Created

Repo: fusengine/agents