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/gmhelmold/hugr_tools/claude-mdgit clone --depth 1 https://github.com/gmhelmold/HuGR_ToolsWrote 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/gmhelmold/hugr_tools/claude-md)<a href="https://agentmods.dev/instructions/gmhelmold/hugr_tools/claude-md"><img src="https://agentmods.dev/badge/instructions/gmhelmold/hugr_tools/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 | $0.00454 | $0.00454 |
| Opus 5 | $0.00227 | $0.00227 |
| Sonnet 5 | $0.00091 | $0.00091 |
| Haiku 4.5 | $0.00045 | $0.00045 |
Grade A, and why
HuGR_Tools 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.
What it actually says
CLAUDE.md — repo working notes for AI agents
Monorepo of 37 MCP toolkit Cloudflare Workers on a shared @hugr/kernel. Remote: HumanGuardrail/HuGR_Tools.
Overview
- Auth: pluggable
IdentityProvider, fail-closed. The kernel resolves its mode strongest-first — RS256 → HMAC → remote introspection (packages/kernel/src/identity/hugr-auth.ts). A self-hoster needs onlyHUGR_INLINE_HMAC_SECRET(local HS256 verification, no network call). Scopes are derived server-side from the caller's plan, never read from the token. - BYOK: workers store no provider key; callers send provider creds as per-request headers; fail-closed.
- CI: GitHub-hosted (
ubuntu-latest);ci.ymlruns the workspace test suite. Reference figures: 8,538 passed / 1,928 credential-gated skips / 0 failed. - Deploy: each package is a standard Cloudflare Worker —
wrangler devlocally, orwrangler deploy --env <env>to your own account. The committedwrangler.tomlvalues (account_id, introspect URL) are placeholders — fill in your own.
Conventions
- Worker names come from each
packages/<pkg>/wrangler.toml[env.<env>].name— NOT alwayshugr-<pkg>-<env>(the Datadog bundles arehugr-datadog-{logs,monitors,dashboards}-<env>). Never derive a worker name from the package name; read the toml. - Secrets never go in the repo or in chat. Real provider tokens live in GitHub Secrets / your machine, not committed.
zodis v4 across the workspace (v3→v4 gotchas: 1-argz.record→ 2-arg,.default({})→.prefault({}), andz.unknown()object fields need explicit.optional()).- Use PRs (squash) off
main; CI must be green.
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 · 18 lines · 454 tokens per session scan A 34ac3293a410
HuGR_Tools CLAUDE.md is an instructions file published in the GitHub repository gmhelmold/HuGR_Tools (0 stars, last pushed 2mo ago), licensed MIT. It adds 454 tokens to every session, about $0.0023 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
mcp-ts-core AGENTS.md
AGENTS.md instructions for cyanheads/mcp-ts-core, covering developer protocol, consumers, core rules, exports reference and import conventions.
aurora AGENTS.md
AGENTS.md instructions for Arvo-AI/aurora, covering aurora - agent guidelines, commands, docker deployments, architecture and secrets management (vault).
projektor AGENTS.md
AGENTS.md instructions for TAJD/projektor, covering agents.md, what projektor is, coordination model (read this first), planning and design docs live in the wiki, not the repo and architecture: the service-layer contract (most important).
senado-br-mcp-cloudflare CLAUDE.md
Claude Code instructions for SidneyBissoli/senado-br-mcp-cloudflare, covering claude.md, project, commands, architecture and testing convention.
ctxlayer CLAUDE.md
Claude Code instructions for stevenn/ctxlayer, covering ctxlayer — claude code briefing, what runs where, architecture pointers, how a claude session should work in this repo and conventions.
context CLAUDE.md
Claude Code instructions for Supa-Media/context, covering agent instructions — context, non-negotiables, architecture, the gateway (apps/mcp) and vocabulary.