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/yarlson/yarstack/agents-mdgit clone --depth 1 https://github.com/yarlson/yarstackWhat 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 | $0.00679 | $0.00679 |
| Opus 5 | $0.00340 | $0.00340 |
| Sonnet 5 | $0.00136 | $0.00136 |
| Haiku 4.5 | $0.00068 | $0.00068 |
Grade A, and why
yarstack 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 yesterday.
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 — 42 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Repository Rules
Commands
- Run
make validatebefore finalizing changes. - Run
make install-system-promptonly when the user explicitly requests changes to global agent guidance; it writes outside the repository. - Test installer changes with
YARSTACK_AGENT_HOMEpointed at a temporary directory. Do not test destructive paths against the real home directory.
Package invariants
- Keep each installable package self-contained under
plugins/<plugin-name>/. Marketplace installs are cached copies; runtime paths must not traverse outside the owning plugin root. - Follow
SKILL_FORMAT.mdwhen adding or changing skills. - Keep Codex metadata in
.codex-plugin/plugin.json, Claude metadata in.claude-plugin/plugin.json, and their marketplace schemas separate. - Keep each plugin's shared identity, author, repository, and version values aligned between its manifests.
- Do not change
0.1.0or any later explicit version unless the user requests a release/version change. When changing it, update both manifests for that plugin together. - Do not add a plugin-root
CLAUDE.md; Claude Code does not load it as plugin context. - Do not declare skills, hooks, agents, MCP servers, apps, assets, or capabilities before the corresponding package content and behavior exist.
Engineering guidance
- Treat
plugins/yarstack/agent-guidance/engineering-standards/as the canonical source for the installed system prompt. - Preserve policy fragment contents and assembly order unless the user explicitly requests prompt changes.
- Preserve installer idempotency, existing symlinks, preflight rejection of dangling symlinks, and backups before replacing differing files.
- Keep guidance installation explicit. Do not convert it into an automatic hook, startup action, or implicit plugin side effect.
Yarbrain invariants
- Keep the canonical vault and all mutable knowledge outside the installed or cached plugin root.
- Do not choose, create, import, or configure a vault during plugin installation. Require an explicit
wiki-initializeaction and preserve existing vault contents. - Keep lifecycle hooks lightweight. They may load a bounded index or enqueue session locators, but must not copy transcript bodies, call a model or network service, or change canonical semantic notes.
- Preserve episodes as evidence. Search before semantic writes, keep proposed changes reviewable, and do not activate a generated skill without explicit approval.
- Test Yarbrain with a temporary
YARBRAIN_CONFIGand vault. Never use the real user configuration or vault.
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.
- yesterday First seen · 42 lines · 679 tokens per session scan A de9d570f8091
yarstack AGENTS.md is an instructions file published in the GitHub repository yarlson/yarstack (3 stars, last pushed 4d ago), licensed MIT. It adds 679 tokens to every session, about $0.0034 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
opensrc AGENTS.md
Instructions for vercel-labs/opensrc, covering agents.md, monorepo structure, cli (packages/opensrc), docs (apps/docs) and releasing.
nanocoder CLAUDE.md
Claude Code instructions for Nano-Collective/nanocoder, covering claude.md, development commands, build and run, testing (run before committing) and individual test commands.
nanocoder AGENTS.md
AGENTS.md instructions for Nano-Collective/nanocoder, covering agents.md, project overview, architecture, key files and development commands.
vibe-coding-repository-standard AGENTS.md
Instructions for cz1993/vibe-coding-repository-standard, covering vcrs repository instructions, purpose, communication, change discipline and privacy and safety.
tokenmaxxing AGENTS.md
Instructions for 851-labs/tokenmaxxing, covering repository instructions, effect error style, export style, conventions and cli output style.
hunch CLAUDE.md
Instructions for davesheffer/hunch, covering claude.md, what this is, commands, architecture and 🧠 hunch (engineering memory).