Marin is an open-source research program, software platform, and community for developing foundation models such as large language models. Researchers use it for data preparation, tokenization, pretraining, posttraining, evaluation, and related experiments, including work on audio-text, DNA, and protein models. The catalogue entries are add-ons that support workflows around Marin.
Borrowing it
Nothing to install: this file belongs to marin-community/marin. 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/marin-community/marin/main/.agents/skills/onboard-marin/SKILL.mdgit clone --depth 1 https://github.com/marin-community/marinWrote 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/skills/marin-community/marin/onboard-marin)<a href="https://agentmods.dev/skills/marin-community/marin/onboard-marin"><img src="https://agentmods.dev/badge/skills/marin-community/marin/onboard-marin/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/skills/marin-community/marin/onboard-marin"><img src="https://agentmods.dev/badge/skills/marin-community/marin/onboard-marin.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 | $0.00062 | $0.00954 |
| Opus 5.5 | $0.00025 | $0.00382 |
| Sonnet 5.5 | $0.00012 | $0.00191 |
| Haiku 4.5 | $0.00006 | $0.00095 |
Grade A, and why
onboard-marin 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 1mo 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 — 85 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Onboard a Marin developer
Establish which parts of the developer environment are ready, fix local setup when safe, and identify each missing external grant. Keep secrets out of command output and the final report.
Sources
Use these as the current sources of truth:
- Local installation:
docs/tutorials/installation.md. - Development workflow:
docs/dev-guide/contributing.md. - Iris authentication and access checks:
lib/iris/OPS.mdand theuse-irisskill. - CoreWeave credentials and routing:
docs/tutorials/cloud-gpu.md. - Pulumi operator grants and state access:
infra/pulumi/README.md, theadd-grantskill, and thereview-grantskill.
Do not copy procedures from infra/README.md; it is an infrastructure index.
Verify
Start with read-only checks. Report each area as ready, missing access, missing local setup, or not checked.
- Confirm the command is running from a Marin checkout and inspect the working tree without changing user work.
- Check the required local tools and Python version. Check dependency and pre-commit setup against the installation and contributing guides. Install or repair local dependencies when the user's onboarding request authorizes it.
- Check GitHub authentication and repository push permission without pushing a branch.
- Check the active GCP account, the
hai-gcp-modelsproject, Application Default Credentials, and read access togs://marin-us-central2. Confirm the current principal hasprojects/hai-gcp-models/roles/marindevwith a filtered IAM query that prints only the matching role name. Do not print credential contents or the complete project policy. - When the user will operate the
marinGCP Pulumi stack:- In addition to the
marindevcheck above, confirm the current principal hasprojects/hai-gcp-models/roles/marinPulumiAdminwith the same filtered IAM query. - Check that the Pulumi CLI and repository deploy dependencies are present.
Run
pulumi -C infra/pulumi stack export --stack marin >/dev/nullto verify state access without printing state contents. - Report CoreWeave kubeconfig access separately when the user will operate a
CoreWeave stack.
Do not run
pulumi upas an onboarding check. Do not grant roles or mutate live IAM unless the user asks for the grant workflow.
- In addition to the
- Check Iris authentication and read-only cluster status. Use
iris loginonly for an interactive human session; let the browser or headless login flow request the human's input. - Check whether
WANDB_API_KEYandHF_TOKENare present without printing their values. When useful, perform a read-only identity check with the service's CLI and confirm access to themarin-communityW&B entity. Do not persist a token outside the user's chosen credential store. - Check CoreWeave object-storage access only when the user needs to inspect GPU job outputs. Do not treat storage credentials as proof of GPU scheduling access; Iris controls compute access separately.
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.
- 1mo ago First seen · 85 lines · 62 tokens per session scan A 84c0159d412d
onboard-marin is a skill published in the GitHub repository marin-community/marin (4,074 stars, last pushed today), licensed Apache-2.0. It adds 62 tokens to every session and 954 once invoked, about $0.0002 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 2026-09-23.
onboard-marin compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| onboard-marinmarin-community/marin | Skill | 4.1k | today | Apache-2.0 | A | 62 | Claude Code, Codex |
| Cloudflaredanielmiessler/LifeOS | Skill | 20k | yesterday | MIT | A | 152 | Claude Code, Codex |
| ccalx-cmd/x-cmd | Skill | 4.7k | 6d ago | Apache-2.0 | A | 10 | Claude Code, Codex |
| x-themex-cmd/x-cmd | Skill | 4.7k | 6d ago | Apache-2.0 | A | 74 | Claude Code, Codex |
| loopx-cmd/x-cmd | Skill | 4.7k | 6d ago | Apache-2.0 | A | 25 | Claude Code, Codex |
| atmos-custom-commandscloudposse/atmos | Skill | 1.4k | yesterday | Apache-2.0 | A | 39 | Claude Code |
| skill0x-cmd/x-cmd | Skill | 4.7k | 6d ago | Apache-2.0 | A | 54 | Claude Code, Codex |