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/drafthq/draft/claude-mdgit clone --depth 1 https://github.com/drafthq/draftWhat 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 | $0.03260 | $0.03260 |
| Opus 5 | $0.01630 | $0.01630 |
| Sonnet 5 | $0.00652 | $0.00652 |
| Haiku 4.5 | $0.00326 | $0.00326 |
Grade A, and why
draft 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 2d 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 — 196 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.
Repository Overview
Draft is a Claude Code plugin that implements Context-Driven Development methodology. It provides a two-tier command surface: 4 primary workflow commands (/draft:init, /draft:new-track, /draft:implement, /draft:review) plus 5 routers (/draft:plan, /draft:ops, /draft:docs, /draft:discover, /draft:jira) as the recommended public interface. 24 specialist commands are dispatched underneath the routers. The unified /draft:jira router supports preview, create, and the advanced review <JIRA-ID> qualification pipeline (deep-review + bughunt + coverage + test-gap analysis). Run /draft for the full intent map. Total surface: 33 skills.
Draft also ships a knowledge graph engine — codebase-memory-mcp, fetched on install to ~/.cache/draft/bin/ (not vendored; see bin/README.md) — driven by scripts/tools/ (53 deterministic shell helpers). Skills are markdown (source of truth, processed by a bash build script into platform-specific integration files for Copilot and Gemini); the graph engine and shell helpers handle mechanical work that markdown can't.
Build & Test Commands
make build # Generate integration files from skills
make build-integrations # Same as above (explicit target)
make test # Run all 80 test suites (skills, build, tools)
make lint # Run shellcheck + markdownlint
make clean # Remove generated integrations
# Run a single test
./tests/test-skill-frontmatter.sh
./tests/test-build-integrations.sh
./tests/test-tools-classify-files.sh
# etc. — any test in tests/ is independently executable
# Graph engine (codebase-memory-mcp — fetched on install, not vendored)
scripts/fetch-memory-engine.sh # install engine to ~/.cache/draft/bin/
scripts/tools/graph-snapshot.sh --repo . # index repo + write draft/graph/schema.yaml gate
scripts/tools/hotspot-rank.sh --repo . # fan-in-ranked hotspots (live query)
scripts/tools/graph-impact.sh --repo . --symbol <name> # blast radius for a symbol (live query)
# Prerequisites: Bash 4.0+, jq (graph tools), Node 18+ (draft CLI), python3 (OKF mode:
# okf-fix-links.sh / okf-render-views.sh), shellcheck, markdownlint-cli (lint only)
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.
- 2d ago First seen · 196 lines · 3,260 tokens per session scan A 2f4a7a41f7e5
draft CLAUDE.md is an instructions file published in the GitHub repository drafthq/draft (40 stars, last pushed 13d ago), licensed MIT. It adds 3,260 tokens to every session, about $0.0163 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 2026-08-30.
Other instructions, from other repositories
memorix CLAUDE.md
Instructions for AVIDS2/memorix, covering memorix - agent instructions for claude code, using memorix memory tools, when to search memory, when to store memory and when to resolve memory.
memorix GEMINI.md
Instructions for AVIDS2/memorix, covering memorix - cross-agent memory rules, session start - bind project, then load context, during session - capture important context, architecture & decisions and bug fixes & problem solving.
Ctxo CLAUDE.md
Instructions for alperhankendi/Ctxo, covering claude.md — ctxo, project overview, quick reference, dev (pnpm workspace) and usage (consumer).
specops CLAUDE.md
Instructions for sanmak/specops, covering claude.md, what is specops, build & validate commands, run all tests (single command) and generate for a single platform.
Ctxo copilot-instructions.md
Instructions for alperhankendi/Ctxo, covering ctxo mcp tool usage (mandatory), before any code modification, before starting a task, before reviewing a pr or diff and when exploring or searching code.
Overture CLAUDE.md
Instructions for SixHq/Overture, covering claude.md, project overview, monorepo structure, build & development commands and install dependencies (from root).