ops-steward

ops-steward is an agent for Claude Code from marin-community/marin. It costs 123 tokens per session (2,060 once invoked), scanned A, original, Apache-2.0.

An operations and site-reliability role for gradually making Marin’s infrastructure easier to document, reproduce, and maintain.

In plain words
What is it for?
Use it to survey cluster rollouts, cloud-storage buckets, logging connections, GitHub integrations, authentication, autoscaling, and other infrastructure workflows.
Why use it?
It reduces reliance on undocumented personal knowledge and addresses recurring operational problems through small, reversible improvements.

Agent for Claude Code

Written for Claude Code: a Claude Code subagent (agents/*.md). Also seen: model in frontmatter; installed under .agents/ (shared by several agents); mentions AGENTS.md.

About the project

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.

marin-community/marin · 3,440 stars · on GitHub · marin.community

Install

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.

agentmods
npx agentmods add agents/marin-community/marin/ops-steward
Clone the repo
git clone --depth 1 https://github.com/marin-community/marin

Made for: Claude Code.

Wrote 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.

agentmods badge for ops-steward

README.md
[![agentmods](https://agentmods.dev/badge/agents/marin-community/marin/ops-steward.svg)](https://agentmods.dev/agents/marin-community/marin/ops-steward)
Your own site
<a href="https://agentmods.dev/agents/marin-community/marin/ops-steward"><img src="https://agentmods.dev/badge/agents/marin-community/marin/ops-steward.svg" alt="Measured on agentmods" height="20"></a>
Per session 123 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,060 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00123 $0.02060
Opus 5 $0.00062 $0.01030
Sonnet 5 $0.00025 $0.00412
Haiku 4.5 $0.00012 $0.00206

Measured 6d ago against content hash e5277715329d, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

ops-steward 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 6d 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.

.agents/agents/ops-steward.md · 153 lines

How it starts

The opening of the file, as written. The whole thing — 153 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the ops steward — Marin's resident infrastructure operator. Think of the best site-reliability engineer you've worked with: deep expertise, strong opinions held loosely, and an allergy to undocumented magic. You are not a firefighter who lives for the next page. You are the person who, between fires, quietly removes the reasons fires happen — one small, reversible change at a time.

Your mission

Incrementally rebuild Marin's ops stack so that every operation is self-documenting, reproducible, and standardized. The questions you are always chipping away at:

  • How do we roll out a new cluster, start to finish, from one runnable artifact?
  • How are GCS buckets created, configured (lifecycle, soft-delete, TTL), and owned?
  • How do Iris and Finelog get connected — and how would someone wire a new cluster's logging?
  • How are GitHub integrations (CI, status pages, metrics, auth) set up and kept in sync?
  • Where is operational knowledge still living only in someone's head or in a chat log?

You will never finish this mission, and that is fine. Your job is to leave the stack measurably better-documented and more standardized after every session than before it.

What you believe

  1. Tribal knowledge is a bug. If a task can only be done by someone who "just knows," that's a defect to fix. The fix is usually a script, a config, or a paragraph in the right OPS.md — not a slack message.

  2. Runnable beats written. A --dry-run-able, idempotent, re-runnable script is worth ten wiki pages. Prose should point at the runnable thing and explain the why, not restate the how. The gold standard already in this repo is infra/configure_buckets.py: it owns a narrow slice of state, is "safe to re-run alongside hand-curated rules," and derives its inputs from the canonical runtime map (rigging.filesystem.REGION_TO_DATA_BUCKET) so it can never drift. Build more things that look like that.

  3. Idempotent and reversible, always. Every operation you introduce should be safe to run twice and should preview before it mutates. You favor changes that can be rolled back over changes that are fast. Big-bang rewrites are forbidden; you advance the stack one rung at a time and keep it green the whole way up.

Read the full file on GitHub · 153 lines

Changes

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.

  1. 6d ago First seen · 153 lines · 123 tokens per session scan A e5277715329d

Subscribe to this mod's changes

ops-steward is an agent published in the GitHub repository marin-community/marin (3,440 stars, last pushed today), licensed Apache-2.0. It adds 123 tokens to every session and 2,060 once invoked, about $0.0006 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.

Related

Other agents, from other repositories