Scion is a multi-agent orchestration platform that runs several isolated agents concurrently in containers, optionally giving each one its own workspace and credentials. It is used to coordinate coding and other research or operations tasks across local machines, remote virtual machines, and Kubernetes clusters. Catalogue add-ons provide workflows and controls for operating Scion.
Borrowing it
Nothing to install: this file belongs to GoogleCloudPlatform/scion. 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/GoogleCloudPlatform/scion/main/AGENTS.mdgit clone --depth 1 https://github.com/GoogleCloudPlatform/scionWrote 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/googlecloudplatform/scion/agents-md)<a href="https://agentmods.dev/instructions/googlecloudplatform/scion/agents-md"><img src="https://agentmods.dev/badge/instructions/googlecloudplatform/scion/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/googlecloudplatform/scion/agents-md"><img src="https://agentmods.dev/badge/instructions/googlecloudplatform/scion/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 | $0.01476 | $0.01476 |
| Opus 5.5 | $0.00590 | $0.00590 |
| Sonnet 5.5 | $0.00295 | $0.00295 |
| Haiku 4.5 | $0.00148 | $0.00148 |
Grade A, and why
scion 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.
How it starts
The opening of the file, as written. The whole thing — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Scion
Alpha: changing or removing behavior needs no deprecation path.
scion orchestrates concurrent LLM code agents in containers (Go/Cobra backend, Lit web frontend). Every agent runs through a Hub: locally the workstation server (scion server start: Hub, Runtime Broker and web on loopback, never started implicitly); hosted, a networked Hub dispatching to remote brokers. Architecture: docs-site/src/content/docs/contributing/architecture.md; design specs: .design/hosted/. "Grove" is the legacy name for "project".
Read before
| Task | Read |
|---|---|
Go build, test or lint in a container; OOM, VCS-status or fmt-check failures |
docs-repo/agents/sandbox-go-toolchain.md |
Code compiled into ./cmd/scion: *_unix.go, private-file writes, path containment, lock files, new commands |
docs-repo/agents/windows-platform-seams.md (CI enforces it) |
Anything under web/, including a new <sl-icon> |
web/AGENTS.md |
| Documentation site | docs-site/AGENTS.md |
| Naming a concept | GLOSSARY.md |
| CLI command agent-mode availability | .design/cli-modes.md |
| Experimental features | docs-site/src/content/docs/reference/experiments.md |
| Workspace or worktree provisioning | docs-site/src/content/docs/local/workspace.md ("Linked git projects") |
Sandbox
- Build with
go build -buildvcs=false; worktrees fail VCS stamping. - Never compile the whole
pkg/hubtest package (about 12 GiB). Usemake test-hub-shard SHARD=K;hack/hubshard/hubshard.sh -locate TestNamefinds the shard. - Scope
golangci-lintto touched packages:GOGC=40 golangci-lint run --new-from-rev=main --concurrency=1 ./pkg/foo/.... - Never open an editor: set the editor per command (
GIT_EDITOR=true git rebase --continue, orgit -c core.editor=true <subcommand>),git merge main --no-edit,git commit -m. Never writecore.editor,core.hooksPathor any other key outside the repository git config allow-list (pkg/util/git_config_policy.go) into the repository config: brokers refuse to start or resume an agent whose checkout sets one. Pass such options per command (git -c core.editor=true <subcommand>). - Local
mainis the source of truth. Don't check it out; usegit diff main...HEAD,git show main:path,git rebase main. Use remotes only when asked; report remote failures instead of working around them. - In worktree-per-agent mode you work under
/repo-root, not/workspace. Scripts use$PWDorgit rev-parse --show-toplevel. - Project resolution:
--project/--global, then the project.scion, thenscion use(outside a project), then~/.scion. - Don't create, delete or touch active agents in this project. Don't commit this repo's
.scion/.
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 Changed · +45 tokens per session b3994d0a64df
- 6d ago Changed · +1 lines · +120 tokens per session e3507a95f173
- 8d ago Changed · -176 lines · -5,262 tokens per session 3dafe91d4374
- 9d ago Changed · +29 lines · +1,269 tokens per session d81b8927e57c
- 14d ago Changed · +64 tokens per session d31a782c774d
- 23d ago Changed · +370 tokens per session af4c2dfa8b43
- 23d ago Changed · +4 tokens per session 5ba24c91bbec
- 25d ago Changed · +1 lines · +176 tokens per session 871273162fff
- 27d ago Changed · +17 tokens per session 3ac3389447a0
- 28d ago First seen · 205 lines · 4,673 tokens per session scan A 2a3baa5b59a7
scion AGENTS.md is an instructions file published in the GitHub repository GoogleCloudPlatform/scion (2,097 stars, last pushed yesterday), licensed Apache-2.0. It adds 1,476 tokens to every session, about $0.0059 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.
scion AGENTS.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| scion AGENTS.mdGoogleCloudPlatform/scion | Instructions file | 2.1k | yesterday | Apache-2.0 | A | 1,476 | Codex, OpenCode |
| vscode buildNext.instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 9,719 | GitHub Copilot |
| next.js AGENTS.mdvercel/next.js | Instructions file | 143k | today | MIT | A | 7,461 | Codex, OpenCode |
| langchain AGENTS.mdlangchain-ai/langchain | Instructions file | 149k | today | MIT | A | 4,680 | Codex, OpenCode |
| vscode oss-third-party-notices.instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 5,516 | GitHub Copilot |
| vscode copilot-instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 3,093 | GitHub Copilot |
| deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness | Instructions file | 274k | 2d ago | MIT | A | 4,313 | Codex, OpenCode |