Provides graph-backed code intelligence and repository mapping for advanced codebase navigation.
Install
mkdir -p .claude/skills/code-explorer && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/12779" && unzip -o skill.zip -d .claude/skills/code-explorer && rm skill.zipInstalls to .claude/skills/code-explorer
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Use Code Explorer / Code Explorer CLI as Codex's code-intelligence layer before broad manual search. Use when answering code questions, mapping repositories such as /home/patrice/code-buddy, tracing symbols, call graphs, impact, dependencies, MCP setup, documentation generation, or when improving TypeScript/JavaScript/C#/Rust graph accuracy.Key capabilities
- →Analyze target repositories for code understanding
- →Check or create code indexes incrementally
- →Query code graphs for specific information
- →Trace symbol usage and dependencies
- →Analyze impact of code changes in both directions
- →Generate documentation for code structure
How it works
This skill indexes code repositories to create a graph-backed map of the code, allowing for deep queries about code structure, symbol usage, and dependencies. It identifies the target repository, checks or creates an index, and then executes graph queries.
Inputs & outputs
When to use code-explorer
- →Map repository structure
- →Trace symbol usage and dependencies
- →Analyze impact of code changes
About this skill
Code Explorer CLI Skill
Use Code Explorer first for repository understanding, then read the exact files it points to before editing or making claims. If graph output conflicts with the source, trust the source and note the graph issue.
You have access to Code Explorer, a code-intelligence tool that builds a knowledge graph from source code. Use it to answer questions about codebases, find symbols, trace dependencies, and analyze impact — using a fraction of the context that reading files would cost.
Binary location
The CLI is code-explorer. After cargo build --release the binary is at target/release/code-explorer; put it on your PATH, or call it by full path from other projects. Inside the Code Explorer repo, cargo run -p code-explorer-cli -- also works.
See the value (quick)
code-explorer demo [path] # Measure the context an LLM agent saves vs reading files
Available commands
1. Analyze a codebase (must run first)
code-explorer analyze [path] # Index a repository
code-explorer analyze [path] --force # Force re-index
code-explorer analyze [path] --incremental # Only re-parse changed files
code-explorer analyze [path] --embeddings # Generate ONNX semantic embeddings (feature gated)
code-explorer analyze [path] --skip-git # Skip git history phases (required for non-git folders)
code-explorer status # Check if index exists (+ last index duration / phase timings)
2. Search the knowledge graph
code-explorer query "authentication middleware" # Natural language search
code-explorer query "user service" --limit 5 # Limit results
3. Symbol context (360-degree view)
code-explorer context UserService # Callers, callees, imports, exports, hierarchy
code-explorer context handleRequest --repo my-project
4. Impact analysis (blast radius)
code-explorer impact handleRequest --direction both # Upstream + downstream
code-explorer impact UserService --direction upstream # Who calls this?
code-explorer impact UserService --direction downstream # What does this call?
5. Ask questions (LLM-powered, requires ~/.codeexplorer/chat-config.json)
code-explorer ask "how does the pricing calculation work?" --path /path/to/project
code-explorer ask "which controllers call the external API?"
6. Code health report
code-explorer report --path /path/to/project # Text report with grade A-E
code-explorer report --path /path/to/project --json # JSON output
7. Git analytics (requires the target to be a git repo)
code-explorer hotspots --path [path] # Most changed files (last 90 days)
code-explorer coupling --path [path] # Files that change together
code-explorer ownership --path [path] # Code ownership by author
8. Tracing coverage & dead code
code-explorer coverage # Global tracing + dead code stats
code-explorer coverage UserService # Single class coverage
code-explorer coverage --json # JSON output
code-explorer coverage UserService --trace # Flow trace mode
9. Raw Cypher queries
code-explorer cypher "MATCH (n:Function) RETURN n.name LIMIT 10"
code-explorer cypher "MATCH (n:Controller)-[:DEFINES]->(a:ControllerAction) RETURN n.name, a.name"
code-explorer cypher "MATCH (n:Method) WHERE n.name STARTS WITH 'Get' RETURN DISTINCT n.name"
code-explorer cypher "MATCH (n:Function) WHERE n.name CONTAINS 'auth' OR n.name CONTAINS 'login' RETURN n"
code-explorer cypher "MATCH (n:Method) WHERE NOT n.filePath ENDS WITH '.test.cs' RETURN n.name"
Supported Cypher operators:
- WHERE:
=,<>,!=,CONTAINS,STARTS WITH,ENDS WITH - Logic:
AND,OR,NOT(precedence: NOT > AND > OR) - RETURN:
DISTINCT,count() - Clauses:
ORDER BY [ASC|DESC],LIMIT
10. Interactive shell
code-explorer shell # REPL with tab completion
# Inside the shell: query, context, impact, cypher, hotspots, stats, help
11. MCP server (for AI agents)
code-explorer mcp # Start MCP server (stdio, JSON-RPC 2.0)
# 30 tools, including: list_repos, query, context, impact, detect_changes, rename,
# cypher, search_code, read_file, find_cycles, find_similar_code, hotspots, coupling,
# ownership, coverage, diagram, report, analyze_execution_trace, get_complexity,
# list_todos, list_endpoints, list_db_tables, list_env_vars, get_endpoint_handler,
# get_insights, save_memory
code-explorer mcp-install # Auto-configure the MCP server for Claude Code
12. HTTP server (REST API)
code-explorer serve --port 3000 # JSON-RPC /mcp + REST + SSE
code-explorer serve --port 3000 --host 0.0.0.0 # Expose on all interfaces
13. Validate LLM config
code-explorer config test # Check API key + test connection
14. List indexed repos
code-explorer list # Show all indexed repositories
15. Generate documentation (all formats)
code-explorer generate context --path [path] # AGENTS.md at repo root
code-explorer generate wiki --path [path] # /wiki/*.md (one per module)
code-explorer generate docs --path [path] # .codeexplorer/docs/*.md
code-explorer generate docx --path [path] # documentation.docx (Word, full doc with TOC)
code-explorer generate html --path [path] # DeepWiki-style HTML site
code-explorer generate obsidian --path [path] # Obsidian vault export
code-explorer generate all --path [path] # All formats above in one run
Flags on every generate subcommand: --output-dir <dir>, --enrich (LLM, needs chat-config.json), --enrich-profile <fast|quality|strict>, --enrich-lang <auto|fr|en>, --enrich-citations.
16. Other commands
code-explorer trace-files ClassName # All source files involved in a feature
code-explorer diagram ClassName --type flowchart|sequence|class
code-explorer rag-import <docs-folder> --path [path] # Import .md/.docx specs, link to code
code-explorer watch [path] # Re-index on file changes (debounced)
code-explorer dashboard [path] # Interactive terminal UI over the graph
code-explorer clean [--force|--all] # Delete an index
code-explorer setup # Configure editor MCP integration
How to use this skill
When the user asks about code structure, architecture, dependencies, or impact:
- Check if the repo is indexed: run
code-explorer statusin the relevant directory. - If not indexed: run
code-explorer analyze [path]first (add--skip-gitif the target isn't a git repo). - Choose the right command based on the question:
- "Where is X defined?" →
code-explorer query "X"orcode-explorer context X - "What calls X?" →
code-explorer impact X --direction upstream - "What does X depend on?" →
code-explorer impact X --direction downstream - "How healthy is the code?" →
code-explorer report - "Which files change most?" →
code-explorer hotspots - "Is this code used?" →
code-explorer coverageorcode-explorer coverage ClassName - "Explain how X works" →
code-explorer ask "how does X work?" - "Show me the architecture" →
code-explorer generate htmlthen read the output - "Which files are involved in X?" →
code-explorer trace-files X - "Generate a diagram of X" →
code-explorer diagram X --type flowchart
- "Where is X defined?" →
- Parse the output and present it clearly with file paths and relationships.
- Combine multiple commands for complex questions.
Graph node types
The knowledge graph contains 50+ node types:
- Code: Function, Method, Class, Interface, Struct, Enum, Module, Namespace
- ASP.NET: Controller, ControllerAction, Service, Repository, ExternalService
- UI: UiComponent (Telerik/Kendo grids), ScriptFile, AjaxCall
- Data: Entity (EF6), Association, NavigationProperty
- Infrastructure: File, Directory, Import, Export
- RAG: Document, DocChunk (from
rag-import)
Relationship types
Calls,CallsAction,CallsService— invocation chainsImports,Exports— module dependenciesInherits,Implements— type hierarchyHasMethod,HasProperty— class membershipBelongsTo,Mentions— RAG doc-to-code anchors
Tips
- The
.codeexplorer/directory in the repo root holds the serialized graph (graph.bin),meta.json, andmetrics.json(index timing/throughput). - Supports 14 languages: JS, TS, Python, Java, C, C++, C#, Go, Rust, Ruby, PHP, Kotlin, Swift, Razor.
- Skips
obj/,bin/,node_modules/,packages/during analysis. - Use
--skip-gitfor non-git folders;hotspots/coupling/ownershipreturn empty there (expected). - All
--jsonflags output machine-readable JSON.
Guardrails
- Do not use Code Explorer output as the sole authority for code edits; verify the exact source.
- Do not run destructive git commands in indexed repositories.
- Use
--skip-gitfor synthetic fixtures and partial source directories. - In a long session, the repository changes under you: re-run
code-explorer analyze --incrementalafter your own commits, and--forcewhen stale output is suspected.
When not to use it
- →Code Explorer output is the sole authority for code edits
- →The task involves destructive git commands in indexed repositories
- →The task requires changing validated C#/ASP.NET/EDMX behavior
Limitations
- →Code Explorer output should not be the sole authority for code edits
- →Destructive git commands should not be run in indexed repositories
- →The `--skip-git` option is used for synthetic fixtures and partial source directories
How it compares
This skill provides a structured, graph-based approach to understanding codebases and their relationships, offering a more detailed and interconnected view than simple file searches or manual code inspection.
Compared to similar skills
code-explorer side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| code-explorer (this skill) | 0 | 3mo | Review | Intermediate |
| solana-dev | 0 | 5mo | No flags | Beginner |
| deepwiki-rs | 25 | 11mo | Review | Intermediate |
| react-component-patterns | 3 | 2mo | No flags | Intermediate |
Try saying
Example prompts that trigger this skill in your AI assistant.
You might also like
solana-dev
solanabr
Unified skill hub for Solana development. Routes to external submodule skills (solana-foundation, sendai, solana-game, trailofbits, cloudflare, qedgen, colosseum) and local skills. Progressive disclosure — read only what you need.
deepwiki-rs
sopaco
AI-powered Rust documentation generation engine for comprehensive codebase analysis, C4 architecture diagrams, and automated technical documentation. Use when Claude needs to analyze source code, understand software architecture, generate technical specs, or create professional documentation from any programming language.
react-component-patterns
HoangNguyen0403
Modern React component architecture and composition patterns.
vercel-architecture-variants
jeremylongshore
Execute choose and implement Vercel validated architecture blueprints for different scales. Use when designing new Vercel integrations, choosing between monolith/service/microservice architectures, or planning migration paths for Vercel applications. Trigger with phrases like "vercel architecture", "vercel blueprint", "how to structure vercel", "vercel project layout", "vercel microservice".
ai-code
arcasilesgroup
Writes production code that satisfies stack-context standards on the first pass: interface-first design, backward-compatibility checks, lightweight self-review. Trigger for 'implement this', 'write the code for', 'add X to Y', 'build this function', 'make this work'. Not for tests; use /ai-test inst
harness-boundary
zkp442910864
Harness 工程边界约束与变更评估。Use when: 评估代码变更影响范围、确认安全边界、检查工程规范合规性、审查变更风险等级。