Search skills/

Understand Anything Claude skills

A 9-skill plugin that maps a codebase, docs or wiki into an interactive knowledge graph with a dashboard, guided tours and Q&A.

Egonex-AI/Understand-Anything 9 skills Planning, TDD & debuggingDocuments & office files
Our verdict Recommended

A well-built, maintained MIT tool with real scripts and an incremental mode; plan for the token cost of the first full run.

4/5 Depth expertise shipped Note Security Review before installing 3d Maintenance Updated 3 days ago 89K Adoption stars OK License MIT, commercial OK

What Understand Anything does

Understand Anything runs a multi-agent pipeline over your project and writes a knowledge-graph.json into a .ua folder: every file, function, class and dependency becomes a node with a plain-English summary. The main /understand skill (a 59 KB SKILL.md with 13 Node and Python helper scripts for scanning, batching, fingerprinting, merging and incremental updates) works through seven phases, reports progress, and later re-runs only touch changed files. Companion skills open the dashboard, answer questions from the graph (/understand-chat), explain a file, analyze a diff, write an onboarding guide, extract business domains and flows, ingest a Karpathy-style markdown wiki, and import Figma. Notes for 10 web frameworks and about 24 file languages sit in subfolders, and output text can be generated in other languages. An optional flag installs a post-commit hook that updates the graph after each commit. Once a graph is committed, teammates can open the dashboard with only Node.js 18 or later, no LLM and no API key; it runs on 127.0.0.1 with a tokenized URL. The README warns that the first run can use many tokens on large repositories.

Good for

  • Onboarding to an unfamiliar or large codebase
  • Teams that want a committed architecture graph everyone can open without an LLM
  • Mapping a markdown knowledge base into a navigable graph

Not for

  • Small projects where reading the code is quicker
  • Token-limited plans without a subscription

What's inside depth 4/5

9 skills averaging 12 KB, the main one with 59 KB of instructions and 13 helper scripts plus language and framework notes, and six eval files in the repo.

Skills
9
Reference docs
0
Scripts
18
Tests / evals
6
Templates
0
Avg SKILL.md
11.6 KB

Counted from the repository file tree. How we score depth.

Dependencies and costs

  • Node.js 18+ for the dashboard and scripts
  • Git (the scanner uses git ls-files)
  • Python for the graph merge scripts
  • Enough LLM tokens for the first full analysis

Security review

Review before installing

A multi-agent codebase-analysis skill that scans your project, writes knowledge-graph files into .ua/ and serves a local dashboard. Install is a curl-to-bash one-liner from the main branch, the dashboard runs a downloaded npx tarball, and the Figma mode uses your FIGMA_TOKEN against Figma. Most of the repo (including plugin manifests and core packages) is missing from our copy, so hooks and install logic could not be checked.

Worth knowing

  • Installs with curl | shREADME install is curl -fsSL https://raw.githubusercontent.com/Egonex-AI/Understand-Anything/main/install.sh | bash (also | bash -s codex / kiro) from the unpinned main branch; install.sh is not in the files we have.README.md:200
  • Relies on a CLI or package we could not reviewThe dashboard skill runs npx --yes on a viewer tarball downloaded from the project's GitHub releases (pinned to the plugin version, not by hash), falling back to pnpm install and npx vite; it serves the knowledge graph on 127.0.0.1 behind a printed token URL.understand-anything-plugin/skills/understand-dashboard/SKILL.md:105
  • Needs an API key for its own serviceThe Figma scan reads FIGMA_TOKEN from the environment and uses it through @understand-anything/core to call the Figma API (core package source is not in our copy).understand-anything-plugin/skills/understand-figma/figma-scan.mjs:14
  • Makes network requestsFetches the Figma file, styles and screen thumbnails from Figma when you run /understand-figma; the dashboard viewer and plugin install download from GitHub.understand-anything-plugin/skills/understand-figma/figma-scan.mjs:14
  • Contains compiled binariesRepo contains two compiled tree-sitter grammars (Dart and Swift .wasm) that cannot be read as source.understand-anything-plugin/packages/tree-sitter-dart-wasm/tree-sitter-dart.wasm:1

What we checked

The repository (521 files) on Oct 29, 2026. Not a full review: 434 of 521 files were not downloaded (large repository; we reviewed the skill folders); it contains compiled binaries or archives we cannot read; we read 4 of 18 code files; it relies on code outside the repository: install.sh fetched by the README one-liner (not in our copy); 434 repo files we do not have, including .claude-plugin/plugin.json, marketplace manifests, the @understand-anything/core and dashboard packages, and any hooks or auto-update logic they define; viewer tarball understand-anything-viewer.tgz downloaded from GitHub releases at runtime.

NVIDIA SkillSpector static scan: 36 pattern hits (Privilege Escalation 11, Tool Misuse 5, Anti-Refusal 5, Data Exfiltration 3, Supply Chain 3, MCP Rug Pull 2). Credential hits are '.env' entries in file-type maps and an access-token URL the localhost dashboard prints, and the Figma token variable; 'rm -rf' hits are cleanup of the skill's own intermediate directory in SKILL.md. README.md line 395 contains an HTML comment addressed to autonomous AI agents asking them to put a marker string in PR descriptions; it is not part of the skill's instructions, but it is text aimed at agents.

A review, not a certification. Skills change after we check them — read the files before installing.

License

MIT — commercial use allowed.

Can it run unattended?

Not really. The analysis runs for a long time on its own and resumes incrementally, but it is a one-off or per-commit job, not a loop that works toward a goal; the optional auto-update hook is the closest to unattended.

Alternatives to Understand Anything

obsidian-wiki

Ar9av/obsidian-wiki

Forty skills that let an agent build and maintain a linked Obsidian vault from your notes, code and past agent sessions, then answer questions with citations.

Recommended Check first Depth 3/5

Archify

tt-a1i/archify

A skill that turns a description or a codebase into a validated, interactive standalone HTML diagram or one-page explainer.

Recommended Check first Depth 5/5

memU

NevaMind-AI/memU

Memory layer that mines agent session logs into a shared markdown wiki and re-injects it across Codex, Claude Code, Cursor and others, locally or via memU Cloud.

With caveats Broad access Depth 4/5

FAQ

What is Understand Anything?

An open-source plugin that analyzes a codebase or knowledge base with a multi-agent pipeline and builds an interactive knowledge graph with a dashboard, guided tours, search and question answering.

How do I install Understand Anything in Claude Code?

Run /plugin marketplace add Egonex-AI/Understand-Anything and then /plugin install understand-anything. For other agents the README offers an install.sh or install.ps1 one-liner, or you can read the script first before running it.

Is Understand Anything safe?

The helper scripts we reviewed only run local git and node commands, and the dashboard binds to localhost with a token. The one thing to check is the curl-to-bash installer offered for non-Claude agents, which we could not inspect.

Does Understand Anything cost a lot of tokens?

The first /understand run analyzes the whole project and the README says it can use a significant number of tokens on large codebases. Later runs are incremental.

Sources: GitHub repository (stars, license, file tree, last push Oct 26, 2026). Review based on the README and SKILL.md files as published.