Guide · Agent skill
How to use the Smart Explore agent skill (smart-explore)
Updated · written with AI from the project's own documentation and checked against our catalogue data
Smart Explore is an agent skill, a set of instructions that tells an AI coding agent how to explore a codebase. Instead of opening whole files, the agent first builds a structural map of functions, classes and methods using tree-sitter syntax parsing, then loads only the specific pieces it needs. This guide is for developers who want to install the skill, ask for it in plain language and understand what it does and does not do.
Quick start
- 1Download and unzip the skill. Download the ZIP from this page and unzip it so that SKILL.md sits inside a smart-explore folder.
- 2Place the folder. Move the folder into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one codebase. Alternatively, upload it under Settings, Capabilities, Skills in the Claude apps.
- 3Confirm the MCP tools exist. Make sure smart_search, smart_outline and smart_unfold are available from the claude-mem plugin, since the skill cannot run searches by itself.
- 4Ask in plain language. Request something like where a function is defined, and name the folder to search. The agent should call smart_search first.
- 5Drill down. Use smart_outline on a file if you need its structure, then smart_unfold on the one symbol you need.
What an agent skill is and what Smart Explore changes
An agent skill is a folder containing a SKILL.md file. The file holds instructions that the AI agent reads and follows when a task matches the skill's description. The skill does not run code on its own.
Smart Explore changes how the agent looks at code. While the skill is active, the source says the agent should use smart_search, smart_outline and smart_unfold as its primary tools instead of Read, Grep and Glob. The core idea is to index first and fetch on demand: get a map of the code before loading implementation details.
- smart_search finds relevant files and ranked symbols across a directory in one call.
- smart_outline returns the structure of one file: functions, classes, methods, properties and imports.
- smart_unfold returns the full source of one named function, class or method.
Install the smart-explore skill
Download the ZIP file for the skill from this page and unzip it. The skill folder contains SKILL.md. Put the unzipped folder in one of these locations, depending on how you want to use it.
You can also upload the skill in the Claude apps. Open Settings, go to Capabilities, then Skills, and add it there.
The skill only gives instructions. The three smart_* tools must come from the wider claude-mem plugin, which this skill's files do not include. Check that those tools are available to your agent before you rely on the skill.
- Personal use: unzip into ~/.claude/skills/
- Single project: unzip into the project's .claude/skills/ folder
- Claude apps: Settings, then Capabilities, then Skills, then upload
Trigger the skill with natural requests
The skill loads when a request matches its description, which covers understanding code structure, finding functions and exploring a codebase efficiently. You do not need special syntax. Ask the way you would ask a colleague.
Good requests name a symbol, a concept or a file. Examples include asking where a function is defined, asking how shutdown is handled in a folder, or asking to show one method from a large file. For exact text searches such as every TODO comment, the skill points the agent to Grep instead.
- Where is performGracefulShutdown defined? Look in ./src.
- Give me an outline of services/worker-service.ts and show the startSessionProcessor method.
- Find everything related to session handling in ./src and show me the key functions.
- Use smart explore to find where the Session class is defined.
Read the results and follow the three-step workflow
A typical run starts with a search. The results list ranked symbols with signatures, line numbers and folded file views. If those views already show enough structure, the agent can skip the outline step and unfold the symbol it needs.
The skill's own examples show the savings it claims. In one example, an outline of a 1,466-token file followed by unfolding one method totals about 3,076 tokens, compared with about 12,000 to read the full file. The skill states 4-8x savings for outline plus unfold versus Read, and 11-18x versus an Explore agent. These are the skill's own figures, not independent benchmarks.
- Search: smart_search(query, path, max_results, file_pattern)
- Outline: smart_outline(file_path)
- Unfold: smart_unfold(file_path, symbol_name)
smart_search(query="shutdown", path="./src", max_results=15)
smart_outline(file_path="services/worker-service.ts")
smart_unfold(file_path="services/worker-service.ts", symbol_name="shutdown")Know which languages and files are supported
Structural analysis works for the bundled language list: JavaScript, TypeScript, TSX and JSX, Python, Go, Rust, Ruby, Java, C and C++. Files with unrecognised extensions do not produce structured symbols for smart_outline or smart_unfold.
smart_search is not a general content search. Files without symbols can only match by path. For JSON, plain config, or any text content search, the skill recommends a text-search tool such as rg. Markdown files get heading-based outlines. Custom grammars cannot yet be added through .claude-mem.json, so installing a grammar in node_modules does not enable parsing.
Safety notes before you use it
The skill runs scripts, according to the assessment for this listing. It does not use the network. Review the code in the claude-mem plugin before installing it in a project that contains sensitive material.
Because the skill overrides the agent's default exploration behaviour while active, check that the agent is using the tools you expect. If the smart_* tools are missing, the agent cannot search, because the skill cannot do it on its own.
What people use it for
- Finding where a function or class is defined in a large TypeScript or Python project.
- Getting the outline of a long file before deciding which method to read.
- Tracing how a concept such as shutdown or session handling is spread across a source folder.
- Reviewing one method in a large Go, Rust or Java codebase without loading every file.
Frequently asked questions
How do I install the smart-explore skill?
Download the ZIP from this page and unzip it. Place the smart-explore folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder, or upload it in the Claude apps under Settings, Capabilities, Skills.
Is smart-explore free to use?
The listing is licensed under Apache-2.0. This page does not state pricing for the wider claude-mem plugin, so check its project for its own terms.
What do I need to run smart-explore?
You need the smart_search, smart_outline and smart_unfold MCP tools, which come from the wider claude-mem plugin rather than the SKILL.md file alone. The skill also depends on tree-sitter parsing for its bundled languages.
Why isn't the skill finding anything in my project?
The most common cause is that the MCP tools are not installed, so the skill has only instructions. Also check that your files use a supported extension, because unsupported files produce no structured symbols.
Can smart-explore search JSON or markdown content?
For JSON and most plain text the skill recommends a text-search tool such as rg. Markdown gets heading-based outlines through smart_outline, but it is not a general content search.
How does smart-explore compare with the Explore agent or Grep?
The skill describes itself as a scalpel for questions like where something is and show me that. For synthesis across six or more files or architecture narratives, it says the Explore agent is better. For exact string or regex matches, it says Grep is the right tool.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Smart Explore suits developers in large JavaScript, TypeScript, Python, Go, Rust, Ruby, Java or C-family codebases who want an agent to find and read specific functions with fewer tokens. It is a poor fit for open-ended architecture questions, non-code files, or any setup where the claude-mem MCP tools are not installed.
Download the skill