AgentSys is a runtime and orchestration system for AI coding agents, coordinating tasks such as branch management, code review, CI, and deployment through structured workflows and persistent state. It is used with Claude Code, Codex CLI, OpenCode, Cursor, and Kiro. The catalogue entries are AgentSys-related plugins, hooks, settings, skills, and instructions.
Borrowing it
Nothing to install: this file belongs to agent-sh/agentsys. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/agent-sh/agentsys/main/AGENTS.mdgit clone --depth 1 https://github.com/agent-sh/agentsysWrote 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/agent-sh/agentsys/agents-md)<a href="https://agentmods.dev/instructions/agent-sh/agentsys/agents-md"><img src="https://agentmods.dev/badge/instructions/agent-sh/agentsys/agents-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/agent-sh/agentsys/agents-md"><img src="https://agentmods.dev/badge/instructions/agent-sh/agentsys/agents-md.svg" alt="Reviewed on agentmods" width="80" 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 | $0.01670 | $0.01670 |
| Opus 5.5 | $0.00668 | $0.00668 |
| Sonnet 5.5 | $0.00334 | $0.00334 |
| Haiku 4.5 | $0.00167 | $0.00167 |
Grade A, and why
agentsys 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 today.
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 — 89 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AgentSys
Marketplace, installer and shared library for the agent-sh plugins. Works on Claude Code, OpenCode, Codex CLI, Cursor and Kiro. Plugins live in their own repos in the agent-sh GitHub org and are pinned by commit in .claude-plugin/marketplace.json; this repo holds the installer (bin/cli.js), the dev CLI (bin/dev-cli.js), lib/ (maintained here; plugin repos vendor their own copies), adapters, docs, checklists and the site.
Critical Rules
- This is a production project with real users: a breaking change reaches every plugin user at once. A feature or fix ships with tests for the changed behavior, edge cases included, and keeps the docs accurate.
- Optimize for developers using the plugins in their own repos, not for internal convenience here.
- Record finished work in CHANGELOG.md and the reply. Summary, audit or completion files (
*_AUDIT.md,*_SUMMARY.md,*_COMPLETION.md) clutter the repo, so leave them out. - Every change reaches main through a PR, whatever its size, including a one-line fix or an urgent hotfix. Never push to main directly. Keep the branch short-lived and rebase it on main before you open the PR. PRs give review, CI and an easy rollback.
- Self-review every PR: a fresh-context subagent reviews the diff and posts its review as a PR comment. Address every comment from revuto (the
revuto-reviewcheck) and every CI failure: fix it, or reply with why not. Merge through the PR once CI is green. When revuto is capped or down, do not wait and do not re-trigger it: the self-review is enough, and the PR body says so. Review bots are not waited on beyond the PR's CI. - Before a multi-file change, read the checklist that applies; multi-file changes have hidden dependencies:
- Cross-platform work:
checklists/cross-platform-compatibility.md(the master reference) - Release:
checklists/release.md - New command, agent, skill or lib module:
checklists/new-command.md,checklists/new-agent.md,checklists/new-skill.md,checklists/new-lib-module.md - OpenCode plugin update:
checklists/update-opencode-plugin.md lib/repo-intelchanges:checklists/repo-intel.md
- Cross-platform work:
- Work is done when every item of that checklist is done. That includes linting new or changed commands, agents, skills, hooks or prompts with agnix (
npx agnix <path>), checking OpenCode and Codex compatibility, and updating thebin/cli.jsmappings for a new command or agent. The pre-push hook runs agnix on the changed agent configs and blocks the push on an error. - Status output uses the plain-text markers
[OK],[ERROR],[WARN],[CRITICAL]and markdown, with no emojis or ASCII-art boxes: they cost tokens and parse worse. - gh and git on Windows: escape
$as\$in GraphQL queries, avoid!=in jq (use== "A" or == "B"), and prefer double quotes with escaped inner quotes over single quotes.gh pr checksreportsstate(SUCCESS,FAILURE,PENDING), notconclusion. The Windows shell treats$and!differently and these fail silently. - Run the pre-commit and pre-push hooks; when one blocks, fix the reported cause and retry.
- Fix every failing test you meet, including ones that look out of scope. A green run has to mean everything works.
- Report script failures before manual fallback. When a project script fails (
npm test,npm run ...,scripts/*,agentsys-dev,node bin/dev-cli.js), report the exact error, find the cause and fix the script. NEVER silently fall back to doing its work by hand: that hides broken tooling.
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.
- today Changed · -14 lines · -204 tokens per session 71372e2b9cd9
- yesterday Changed · -5 lines · -124 tokens per session 65b92f8c3d59
- 13d ago Changed · +1 lines · +25 tokens per session 36bcf61ae3c3
- 21d ago Changed · +49 tokens per session 8ced68203127
- 22d ago Changed · -197 lines · -1,539 tokens per session 41263df68d9e
- 1mo ago Changed · +26 lines · +320 tokens per session 5563cb318466
- 1mo ago Changed · +28 tokens per session ee0bcc7a2f27
- 1mo ago First seen · 278 lines · 3,115 tokens per session scan A 24a2c0b740bd
agentsys AGENTS.md is an instructions file published in the GitHub repository agent-sh/agentsys (1,009 stars, last pushed yesterday), licensed MIT. It adds 1,670 tokens to every session, about $0.0067 per session on Opus 5.5. 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
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).
deno copilot-instructions.md
Copilot instructions for denoland/deno, covering deno development guide for github copilot, network access, high level overview, key directories and key files to understand first.