Skip to content
FunCoding

Search

Search docs, Skills and MCP

explore

Meta-skill for internal codebase exploration at varying depths (quick/deep/architecture)

代码质量与审查3.9k.claude/skills/explore/SKILL.md

Install

Send this to Claude Code, Codex or Cursor. The agent checks the Skill for safety first and installs it only after you confirm.

读取 https://funcoding.ai/skills/parcadei/continuous-claude-v3/explore/install.md ,按里面的步骤帮我安装这个 Skill。

SKILL.md

Explore - Internal Codebase Exploration

Meta-skill for exploring an internal codebase at varying depths. READ-ONLY workflow - no code changes.

Usage

/explore <depth> [options]

Question Flow (No Arguments)

If the user types just /explore with no or partial arguments, guide them through this question flow. Use AskUserQuestion for each phase.

Phase 0: Workflow Selection

question: "How would you like to explore?"
header: "Explore"
options:
  - label: "Help me choose (Recommended)"
    description: "I'll ask questions to pick the right exploration depth"
  - label: "Quick - fast overview"
    description: "Chain: tldr tree → tldr structure (~1 min)"
  - label: "Deep - comprehensive analysis"
    description: "Chain: onboard → tldr → research → document (~5 min)"
  - label: "Architecture - layers & dependencies"
    description: "Chain: tldr arch → call graph → layer mapping (~3 min)"

Mapping:

  • "Help me choose" → Continue to Phase 1-4 questions
  • "Quick" → Set depth=quick, skip to Phase 2 (scope)
  • "Deep" → Set depth=deep, skip to Phase 2 (scope)
  • "Architecture" → Set depth=architecture, skip to Phase 2 (scope)

If Answer is Unclear (via "Other"):

question: "I want to understand how deep you want to explore. Did you mean..."
header: "Clarify"
options:
  - label: "Help me choose"
    description: "Not sure - guide me through questions"
  - label: "Quick - fast overview"
    description: "Just want to see what's here"
  - label: "Deep - comprehensive analysis"
    description: "Need thorough understanding"
  - label: "Neither - let me explain differently"
    description: "I'll describe what I need"

Phase 1: Exploration Goal

question: "What are you trying to understand?"
header: "Goal"
options:
  - label: "Get oriented in the codebase"
    description: "Quick overview of structure"
  - label: "Understand how something works"
    description: "Deep dive into specific area"
  - label: "Map the architecture"
    description: "Layers, dependencies, patterns"
  - label: "Find where something is"
    description: "Locate specific code/functionality"

Mapping:

  • "Get oriented" → quick depth
  • "Understand how" → deep depth
  • "Map architecture" → architecture depth
  • "Find where" → quick with --focus

Phase 2: Scope

question: "What area should I focus on?"
header: "Focus"
options:
  - label: "Entire codebase"
    description: "Explore everything"
  - label: "Specific directory or module"
    description: "I'll specify the path"
  - label: "Specific concept/feature"
    description: "e.g., 'authentication', 'API routes'"

If "Specific directory" or "Specific concept" → ask follow-up for the path/keyword.

Phase 3: Output Format

question: "What should I produce?"
header: "Output"
options:
  - label: "Just tell me what you find"
    description: "Interactive summary in chat"
  - label: "Create a documentation file"
    description: "Write to thoughts/shared/docs/"
  - label: "Create handoff for implementation"
    description: "Prepare context for coding agent"

Mapping:

  • "Documentation file" → --output doc
  • "Handoff for implementation" → --output handoff

Phase 4: Entry Point (Architecture only)

If architecture depth selected:

question: "Where should I start the analysis?"
header: "Entry point"
options:
  - label: "Auto-detect (main, cli, app)"
    description: "Find common entry points"
  - label: "Specific function/file"
    description: "I'll specify the entry point"

Summary Before Execution

Based on your answers, I'll run:

**Depth:** deep
**Focus:** "authentication"
**Output:** handoff
**Path:** src/

Proceed? [Yes / Adjust settings]

Depths

DepthTimeWhat it does
quick~1 mintldr-explorer only - fast structure overview
deep~5 minonboard + tldr-explorer + research-codebase + write doc
architecture~3 mintldr arch + call graph + layer mapping + circular dep detection

Options

OptionDescriptionExample
--focus "area"Focus on specific area--focus "auth", --focus "api"
--output handoffCreate handoff for next agent--output handoff
--output docCreate documentation file--output doc
--entry "func"Start from specific entry point--entry "main", --entry "process_request"

Examples

# Quick structure overview
/explore quick

# Deep exploration focused on auth
/explore deep --focus "auth" --output doc

# Architecture analysis from specific entry
/explore architecture --entry "cli" --output handoff

# Quick focused exploration
/explore quick --focus "hooks"

Workflow Details

Quick Depth

Fast structure overview using tldr-explorer. Best for:

  • Initial orientation
  • Quick questions about structure
  • Finding where things are

Steps:

  1. Run tldr tree for file structure
  2. Run tldr structure for codemaps
  3. If --focus provided, run tldr search for targeted results
  4. Return summary

Commands:

# 1. File tree
tldr tree ${PATH:-src/} --ext .py

# 2. Code structure
tldr structure ${PATH:-src/} --lang python

# 3. Focused search (if --focus provided)
tldr search "${FOCUS}" ${PATH:-src/}

Deep Depth

Comprehensive exploration with documentation output. Best for:

  • First time in a codebase
  • Preparing for major work
  • Creating reference documentation

Steps:

  1. Check if onboarded (look for .claude/cache/tldr/), if not run onboard
  2. Run tldr-explorer for structure
  3. Spawn research-codebase agent for patterns
  4. Write findings to doc or handoff

Subprocess:

# 1. Onboard check
if [ ! -f .claude/cache/tldr/arch.json ]; then
    # Spawn onboard agent
fi

# 2. Structure analysis
tldr structure src/ --lang python
tldr calls src/

# 3. Research patterns (via scout agent)
Task: research-codebase → "Document existing patterns in ${FOCUS:-codebase}"

# 4. Write output
→ thoughts/shared/research/YYYY-MM-DD-explore-{focus}.md
→ OR thoughts/shared/handoffs/{session}/explore-{focus}.yaml

Architecture Depth

Architecture-focused analysis with layer detection. Best for:

  • Understanding system boundaries
  • Preparing for refactoring
  • Identifying coupling issues

Steps:

  1. Run tldr arch for layer detection
  2. Run tldr calls for cross-file call graph
  3. Analyze entry/middle/leaf layers
  4. Detect circular dependencies
  5. Map architectural boundaries

Commands:

# 1. Architecture detection
tldr arch ${PATH:-src/}
# Returns: entry_layer, middle_layer, leaf_layer, circular_deps

# 2. Call graph
tldr calls ${PATH:-src/}
# Returns: edges, nodes

# 3. Impact analysis from entry point (if --entry provided)
tldr impact ${ENTRY} ${PATH:-src/} --depth 3

Output Structure:

layers:
  entry: [routes.py, cli.py, main.py]  # Controllers/handlers
  middle: [services.py, auth.py]        # Business logic
  leaf: [utils.py, helpers.py]          # Utilities

call_graph:
  total_edges: 142
  hot_paths: [process_request → validate → authorize]

circular_deps:
  - [module_a, module_b]  # A imports B, B imports A

boundaries:
  - name: API layer
    files: [src/api/*]
    calls_to: [src/services/*]

Output Formats

--output doc

Creates: thoughts/shared/research/YYYY-MM-DD-explore-{focus}.md

---
date: {ISO timestamp}
type: exploration
depth: {quick|deep|architecture}
focus: {focus area or "full"}
commit: {git hash}
---

# Codebase Exploration: {focus}

## Summary
{High-level findings}

## Structure
{File tree / codemaps}

## Architecture
{Layer analysis - for architecture depth}

## Key Components
{Important files and their roles}

## Patterns Found
{Existing patterns - for deep depth}

## References
- `path/to/file.py:line` - Description

--output handoff

Creates: thoughts/shared/handoffs/{session}/explore-{focus}.yaml

---
type: exploration
ts: {ISO timestamp}
depth: {quick|deep|architecture}
focus: {focus area}
commit: {git hash}
---

summary: {One-line summary of findings}

structure:
  entry_points: [{main.py}, {cli.py}]
  key_modules: [{auth.py}, {routes.py}]
  test_coverage: [{tests/}]

architecture:
  layers:
    entry: [{files}]
    middle: [{files}]
    leaf: [{files}]
  circular_deps: [{pairs}]

findings:
  - {key finding with file:line}

next_steps:
  - {Recommended action based on exploration}

refs:
  - path: {file.py}
    role: {what it does}

Integration with /build

The explore skill is designed to feed into /build brownfield:

# Step 1: Explore to understand
/explore architecture --output handoff

# Step 2: Build with context from exploration
/build brownfield --from-handoff thoughts/shared/handoffs/session/explore-full.yaml

Implementation

When user invokes /explore <depth> [options]:

Parse Arguments

depth = args[0]  # quick | deep | architecture
focus = extract_option(args, "--focus")
output = extract_option(args, "--output")  # handoff | doc
entry = extract_option(args, "--entry")

Execute Based on Depth

Quick:

# Just tldr commands, no agents
tldr tree ${src_dir} --ext .py
tldr structure ${src_dir} --lang python
if [ -n "$focus" ]; then
    tldr search "$focus" ${src_dir}
fi

Deep:

# 1. Check/run onboard
if [ ! -f .claude/cache/tldr/meta.json ]; then
    # Spawn onboard agent via Task tool
fi

# 2. Structure
tldr structure src/ --lang python

# 3. Research (spawn scout agent)
# Task tool with subagent_type: "scout"
# Prompt: "Research patterns in ${focus:-codebase}"

# 4. Write output
# → doc or handoff based on --output

Architecture:

# 1. Arch detection
arch_output=$(tldr arch ${src_dir})

# 2. Call graph
calls_output=$(tldr calls ${src_dir})

# 3. Impact from entry (if provided)
if [ -n "$entry" ]; then
    impact_output=$(tldr impact $entry ${src_dir} --depth 3)
fi

# 4. Synthesize and write output

Key Principles

  1. READ-ONLY - This skill never modifies code
  2. Uses scout, not Explore - Per project rules, scout (Sonnet) over Explore (Haiku)
  3. Token-efficient - Uses tldr commands (95% savings over raw reads)
  4. Outputs to shared locations - thoughts/shared/research/ or handoff directory
  5. Entry point to /build - Exploration handoffs feed into brownfield builds
SkillWhen to Use
tldr-explorerDirect tldr commands (used internally by explore)
tldr-codeSpecific analysis commands (cfg, dfg, slice)
onboardFirst-time project setup (used by deep depth)
research-codebasePattern documentation (used by deep depth)
create_handoffHandoff format (used by --output handoff)

Troubleshooting

tldr not found:

# Check if installed
which tldr
# Install if missing
uv tool install llm-tldr
# or: pip install llm-tldr

No Python files found:

# Check language, adjust --lang
tldr structure src/ --lang typescript  # or go, rust

Empty architecture output:

# May need to specify src directory
tldr arch ./  # Current directory
tldr arch src/  # Explicit src

Similar Skills

claude-api
anthropics/skills180k

claude-api

Reference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file).

Code quality & review

ponytail-review
DietrichGebert/ponytail158k

ponytail-review

Quality review of a change: is the logic right, is it safe, does it hold under real load, is risky code tested, is it fast enough, and is every line needed. Reads the connected code, not only the diff. Each finding is explained in plain English. Use for "review this", "code review", "review the last commit", "review my PR", "is this over-engineered", /ponytail-review.

Code quality & review

code-review-and-quality
addyosmani/agent-skills103k

code-review-and-quality

Conducts multi-axis code review. Use before merging any change. Use when reviewing code written by yourself, another agent, or a human. Use when you need to assess code quality across multiple dimensions before it enters the main branch. Use when asked to review a diff or a pull request, even when the diff is pasted inline.

Code quality & review

documentation-and-adrs
addyosmani/agent-skills103k

documentation-and-adrs

Records decisions and documentation. Use when you need to document an architecture decision (ADR) or the reasoning behind a design choice, when changing public APIs, shipping features, or when you need to record context that future engineers and agents will need to understand the codebase.

Code quality & review

code-simplification
addyosmani/agent-skills103k

code-simplification

Simplifies code for clarity. Use when refactoring code for clarity without changing behavior. Use when code works but is harder to read, maintain, or extend than it should be. Use when reviewing code that has accumulated unnecessary complexity.

Code quality & review

understand
Egonex-AI/Understand-Anything86k

understand

Analyze a codebase to produce an interactive knowledge graph for understanding architecture, components, and relationships

Code quality & review