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/damusix/atomic-claude/claude-mdgit clone --depth 1 https://github.com/damusix/atomic-claudeWrote 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/damusix/atomic-claude/claude-md)<a href="https://agentmods.dev/instructions/damusix/atomic-claude/claude-md"><img src="https://agentmods.dev/badge/instructions/damusix/atomic-claude/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 5.1 | $0.08859 | $0.08859 |
| Opus 5 | $0.04430 | $0.04430 |
| Sonnet 5 | $0.01772 | $0.01772 |
| Haiku 4.5 | $0.00886 | $0.00886 |
Grade C, and why
atomic-claude CLAUDE.md scanned grade C with 2 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 6d 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.
Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
Install like any user: `curl -fsSL https://raw.githubusercontent.com/damusix/atomic-claude/main/install.sh | bash` fetches the binary, then `atomic claude install` writes the artifact bundle into `~/.claude/`. To test lo Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
Install like any user: `curl -fsSL https://raw.githubusercontent.com/damusix/atomic-claude/main/install.sh | bash` fetches the binary, then `atomic claude install` writes the artifact bundle into `~/.claude/`. To test lo How it starts
The opening of the file, as written. The whole thing — 365 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
Project-local context for working on this repo. Not copied anywhere — read by Claude only when the cwd is this repo. The contract that ships to users is context/CLAUDE.md, a different file.
What this repo is
A holistic Claude Code configuration. Everything that installs lives under context/: CLAUDE.md, commands/, agents/, skills/, rules/, output-styles/. They are designed as one coherent system — atomic output style, an opinionated command set, a small subagent roster, and discipline skills that interlock. Not a grab-bag; everything is meant to compose.
Replaces (for the author) heavier toolkits like superpowers and caveman. Personal config, no stability guarantee.
Platform support
Target macOS and Linux only. Drop Windows-specific review findings, Windows-only test paths (os.PathSeparator probes), and Windows compatibility gates. .goreleaser.yaml may still produce Windows binaries — that's fine, but correctness on Windows is not a concern.
File roles (this repo specifically)
| File | Role | Destination |
|---|---|---|
context/CLAUDE.md |
Single source of truth. Two roles in one file: (a) the global contract that ships as every user's ~/.claude/CLAUDE.md on install, and (b) this repo's own project instructions when working on atomic-claude. Since it no longer sits at the repo root, Claude Code does not auto-load it here — the @context/CLAUDE.md ref below pulls it in. Not to be confused with the root CLAUDE.md, which is the row below. |
context/, committed → ~/.claude/CLAUDE.md on install |
CLAUDE.md (this file) |
Project-local overlay for this repo only. Build pipeline rules, doc reference paths, mandatory checklist, design axioms. Auto-loads at the repo root, and pulls in context/CLAUDE.md by @-ref. Do NOT duplicate context/CLAUDE.md content here — both load into context, duplication = wasted tokens. |
Repo root, committed, never installed. |
README.md |
Human-facing overview of what the config does and how to install it. | Repo root, committed. |
context/commands/*.md |
Slash command definitions, committed in source form. May carry {{ template "<flow>" . }} directives resolved against context/_partials/; expansion happens on the way into the embedded bundle, never back into context/. Copied to ~/.claude/commands/ by atomic claude install. |
~/.claude/commands/ |
context/agents/*.md |
Subagent definitions, same contract as commands. Every agent composes at least agent-atomic-voice. |
~/.claude/agents/ |
context/_partials/<name>.md |
Reusable blocks composed by command AND agent sources via {{ template "<name>" . }}. One shared pool — a partial defined once is callable from either kind. Never installs: the mirror walks only the artifact kind directories, and _partials/ is not one of them. The full inventory and the per-artifact composition tables live in docs/wiki/bundle.md — read them there rather than keeping a second copy here, which is what let this row drift. One rule that is not derivable from the files themselves: agent-yagni is kept verbatim in sync with the Simplicity-first (YAGNI) ladder in context/CLAUDE.md's Principles block, and CLAUDE.md is not expanded, so nothing enforces the match. |
Not copied; consumed at build time. |
context/skills/*/SKILL.md |
Discipline skills. Copied to ~/.claude/skills/. |
~/.claude/skills/ |
context/output-styles/*.md |
Output style definitions. Copied to ~/.claude/output-styles/. |
~/.claude/output-styles/ |
context/rules/<topic>/*.md |
Shipped path-scoped topic rules. paths: frontmatter globs (e.g. **/*.{ts,tsx}, docs/spec/**) so the rule only loads when Claude touches a matching file — auto-loads into subagents too (verified). Currently: typescript/, python/ (language style), specs/spec-currency.md (spec body-is-truth, globs docs/spec/** + docs/design/**). |
~/.claude/rules/ (via atomic claude install) |
.claude/rules/<topic>/*.md |
Repo-only path-scoped rules — committed (gitignore-negated) but NOT bundled, so they never ship to users. .claude/rules/authoring/ holds the contributor artifact-authoring references (agent-config, prompting, claude-code-refs, axioms) — glob the artifact sources (context/**) so they auto-load only when editing an artifact, instead of @-ref'ing every session. |
Stays in repo; never installed. |
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.
- 6d ago First seen · 365 lines · 8,859 tokens per session scan C dadce7140580
atomic-claude CLAUDE.md is an instructions file published in the GitHub repository damusix/atomic-claude (84 stars, last pushed yesterday), licensed MIT. It adds 8,859 tokens to every session, about $0.0443 per session on Opus 5. A static security scan graded it C with 2 findings (downloads and executes remote code, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
crystal CLAUDE.md
Instructions for stravu/crystal, covering crystal - multi-session claude code manager, project overview, references, implementation status: ✅ complete and ✅ implemented features.
easy-cheese python.instructions.md
Instructions for paulnsorensen/easy-cheese, covering hard rules, what to flag and what not to flag.
plan-forge database.instructions.md
Database patterns for .NET — Dapper/EF Core, parameterized queries, migration strategy.
plan-forge naming.instructions.md
Azure naming conventions — Microsoft CAF abbreviations, constraints, uniqueness, tagging.
claude-bridge AGENTS.md
Instructions for kadaba/claude-bridge, covering agents.md — using claude-bridge from any ai coding cli, two independent channels, 1. coordination (the conversation), 2. file transfer (the bytes) — same for every tool and workflows.
AmbyKit CLAUDE.md
Claude Code instructions for ambystechcom/AmbyKit, covering claude code — notes for the ambykit repo and claude-specific.