Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/adityaarakeri/gitatlas/claude-mdgit clone --depth 1 https://github.com/adityaarakeri/gitatlasWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/adityaarakeri/gitatlas/claude-md)<a href="https://agentmods.dev/instructions/adityaarakeri/gitatlas/claude-md"><img src="https://agentmods.dev/badge/instructions/adityaarakeri/gitatlas/claude-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.01685 | $0.01685 |
| Opus 5 | $0.00843 | $0.00843 |
| Sonnet 5 | $0.00337 | $0.00337 |
| Haiku 4.5 | $0.00169 | $0.00169 |
Grade A, and why
gitatlas CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 79 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
What this is
gitatlas turns a folder of git repositories into one self-contained interactive HTML architecture map (group -> repo -> module -> symbol). No server, no database, no network at runtime. Everything is local files.
Commands
npm install # no build step, no native deps; code runs through tsx directly
npm test # full suite: extractor (all 21 languages), analysis, scoper (Node built-in test runner)
npm run extract -- fixtures # smoke test: indexes 3+ fixture repos, emits fixtures/.gitatlas/index.html
npm run check -- fixtures # freshness check against the existing map: exit 0 fresh, 1 stale; --json for machines
npm run extract -- fixtures --if-stale # re-extract only when source content changed (run before trusting graph data)
npm run brief -- --out fixtures/.gitatlas --budget 1500 # token-budgeted markdown digest for agent context
npm run mcp -- --out fixtures/.gitatlas # MCP server over stdio (brief/scope/find_symbol/module_info/check_freshness)
npm run scope -- --trace crash.txt --out fixtures/.gitatlas # bug scoper (requires a prior extract)
Run a single test file:
npm test -- packages/extractor/test/extract.test.ts
scripts/test.mjs wires the flags for both: single-process isolation (child test processes crash compiling tree-sitter grammars on Node 24, nodejs/node#63421) plus --liftoff-only on Node 24+. The isolation flag is detected, not hardcoded: Node 22 spells it --experimental-test-isolation, Node 24 spells it --test-isolation, and the wrong spelling is a fatal node: bad option at startup.
Requires Node 22+ (22.8+ for the test suite).
Architecture
Data flows one way: extractor -> JSON graph files -> viewer HTML. The pieces communicate only through the schema.
packages/schema/src/types.ts— the contract.RepoGraph+GroupManifestJSON shapes,SCHEMA_VERSION. Start here for any change.packages/extractor/— CLI (cli.ts): repo discovery (scans for.gitmarkers), TS/JS extraction via the TypeScript compiler API, viewer generation (embeds data + vendored d3 into the template).github.tsparses GitHub targets including the ref a/tree/<ref>/URL carries (pure, shared withpackages/site) andclone.tsresolves a target to a local directory, shallow-cloning into the cache when it names a GitHub repo. One clone per (owner, repo, ref); the clone's basename is the repo name the map carries.extract.tsholds the tree-sitter WASM engine andSITTER_LANGS, a per-language config table covering the other 19 languages.layout.tscomputes node positions at extract time.packages/analysis/— neighborhoods (single-level Louvain clustering) and hub detection (degree statistics).packages/scoper/— resolves a stack trace or symbol name into the graph and ranks the surrounding neighborhood.scope.tsis pure functions,cli.tsis I/O only; mirror that split for new logic.packages/brief/— token-budgeted markdown digest of the map for agent context (brief.tspure,cli.tsI/O). Budget cuts are always announced in the text, never silent.packages/mcp/— dependency-free MCP server over stdio (server.tstool logic,cli.tsJSON-RPC framing). Re-verifies source fingerprints before every tool call and flags stale repos inside each result.packages/viewer/template.html— the entire viewer: one file, vanilla JS + d3, no framework, no build. Keep it view-source-able.packages/viewer/locales/<code>.jsonholds the UI catalogs the extractor bakes in;en.jsonis the source of truth anddocs/i18n.mdis the contributor guide.bin/gitatlas.js— global entry; re-execs with--liftoff-onlyon Node 24+, then dispatches to extractor or scoper CLI viatsx/cjs.fixtures/— small fake repos used by tests and the smoke test;fixtures/polyglot-labhas one file per tree-sitter language.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 79 lines · 1,685 tokens per session scan A cb7a74c5d0d8
gitatlas CLAUDE.md is an instructions file published in the GitHub repository adityaarakeri/gitatlas (0 stars, last pushed 14d ago), licensed MIT. It adds 1,685 tokens to every session, about $0.0084 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
deepseek-harness AGENTS.md
AGENTS.md instructions for deepseek-ai/deepseek-harness, covering agents.md, pre-release stance: foundation over blast radius, repository layout, commands and host sandbox failures.