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/royashbrook/hush/agents-mdgit clone --depth 1 https://github.com/royashbrook/hushWhat 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 | $0.00410 | $0.00410 |
| Opus 5 | $0.00205 | $0.00205 |
| Sonnet 5 | $0.00082 | $0.00082 |
| Haiku 4.5 | $0.00041 | $0.00041 |
Grade A, and why
hush AGENTS.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.
What it actually says
AGENTS.md, for any agent, not just one tool
You were probably pointed here as "use the hush skill." This file is the agent-agnostic entry point;
the full playbook is SKILL.md.
what hush is (one line)
A secret store you operate so the user never has to paste a secret into the chat: get a value once into the OS keychain, then inject it into the already-authed CLIs you're running as the user, forever, never printed, never written to the repo.
install (any agent)
hush is a single self-contained bash script. Get it on the box and make it executable:
git clone https://github.com/royashbrook/hush
chmod +x hush/hush
Then either put hush/hush on your PATH, or move it into wherever your agent loads its
tools/skills. That's all the install there is, no build, no deps beyond your OS secret backend
(macOS Keychain / Linux secret-tool / Windows DPAPI, all auto-detected).
Tool-specific convenience:
- Claude Code, clone into
~/.claude/skills/hush(whole session) or./.claude/skills/hush(one project) and it auto-loads as a skill. - Anything else, PATH or your tool's plugin/skill directory works the same; hush is just a script.
then read SKILL.md
SKILL.md is the contract and the playbook: the two add-paths (hush set pops a hidden
dialog for a value the user holds; hush mint generates a random one), injecting with hush run /
hush pipe, adopting hush in an existing repo, and the one hard rule, you never see the
plaintext, and there is no get. Follow it directly.
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 · 36 lines · 410 tokens per session scan A 93d7194c674f
hush AGENTS.md is an instructions file published in the GitHub repository royashbrook/hush (20 stars, last pushed 19d ago), licensed MIT. It adds 410 tokens to every session, about $0.0020 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
latchkey AGENTS.md
Instructions for imbue-ai/latchkey: For information about the high-level goal and motivations, see README.md.
treg CLAUDE.md
Claude Code instructions for superdesigndev/treg, covering tools-registry (treg), words to use, and words not to, do not document what is not built, repo conventions and development.
treg AGENTS.md
AGENTS.md instructions for superdesigndev/treg, covering agents.md — guide for ai coding agents, read first, working agreement, do not touch (without reading the fragment first) and security awareness.
kdbx AGENTS.md
Instructions for yarrasys/kdbx, covering agents.md, what this is, current status — read this first, golden rules and repository map.
xaip-protocol AGENTS.md
AGENTS.md instructions for xkumakichi/xaip-protocol, covering agents.md — ai-assisted contribution guide, 1. public wording boundaries, 2. commit and pr rules, 3. privacy rules and 4. verification before commit.
agentvault CLAUDE.md
Claude Code instructions for bch1212/agentvault, covering agentvault, what it does, stack, project structure and running locally.