Sno Station Skills

What does this agent do, and when should you use it?

Sno Station Skills is the skill set that Sno Station installs into Claude Code and Codex. The badges and body describe 37 skills shipped as 28 units, grouped into six team capabilities (J peer review, M build, S communication, H reporting, T delegation, R retrospective). Every SKILL.md opens with a requires block naming the Sno Station programs it calls (reach, heartbeat, subscription-quota-check, report-time) and the harness slots it needs, so sno setup can install it fully, in reduced form, or skip it per agent. The core pattern is Dual Brain: one agent does the work, an agent from the other vendor reviews it in a fresh session with none of the author's conversation, and the author fixes findings until the review comes back clean. Before a long run you can arm rotate-agent, which watches remaining quota and hands the task to the other vendor's agent with a verified handover brief and the exact commit to continue from. Work counts as done only when a written list of success checks is answered by real runs. Skills install via sno setup or the Claude Code and Codex plugin marketplaces, and the repository is Apache-2.0.

Skills are Markdown instructions plus scripts that the agent executes. peer-review launches the other vendor's CLI in a new session against a diff, plan or document and drives a fix loop; less-is-more climbs a ladder (needed at all, already in the repo, standard library, platform, installed dependency) before writing code and then trims its own diff; codex-coder keeps the plan in Claude Code and runs codex exec for the coding, killing and restarting a run that has written nothing for 15 minutes; rotate-agent uses heartbeat to read subscription-quota-check every few minutes, then triggers handoff to write a brief and wake the other agent; deliver answers each charter success check with a recorded run tracked by sno deliver-proof; rem-reflect reads the day's local Claude Code and Codex sessions and proposes lessons and skill changes that only apply after a human runs sno rem-reflect accept; away-brief emits a page of at most 60 lines split into Done, Stuck, Needs you and Spend; catch-report counts problems caught by the other model separately from self-caught mistakes, never summing them.

1. A developer finished a change and wants a reviewer from a different model vendor to read it cold and report what would break in real use.
2. A long build runs overnight and the team wants the task to survive one subscription running out of quota by handing over to the other vendor's agent.
3. A user returns in the morning and needs one page covering what got done, what is stuck, which decisions are waiting and how much quota was spent.
4. A lead must delegate a job to multiple executor agents and needs a written charter with deliverables, out-of-bounds areas and success checks.
5. A team wants agents to learn from repeated corrections but insists that every lesson be approved by a human before it changes a skill.
6. A user wants to try a single skill first, such as less-is-more or test-writer, which run without Sno Station.

How do you install or deploy this agent?

Ask your agent in chat to install sno.ai, or run the install script:

curl -fsSL https://sno.ai/install | sh

If Sno Station is already present, run sno setup to install the skills into your agents; run it again when you add another agent.

You can also install as a plugin. Claude Code:

claude plugin marketplace add sno-ai/sno-station-skills
claude plugin install sno-station@sno-ai

Codex:

codex plugin marketplace add sno-ai/sno-station-skills
codex plugin add sno-station@sno-ai

For a single skill, have the agent install it from the repository, for example peer-review:

Install only the peer-review skill from https://github.com/sno-ai/sno-station-skills
for this project, including its scripts and references.
Check its dependencies and tell me what is missing before installing anything else.

rotate-agent additionally needs both CLIs logged in and tmux installed.

How do you use this agent?

After installation, trigger skills in ordinary chat, for example "peer-review this", "arm rotate-agent", "away brief" or "write a charter". The sno-station plugin holds all 37 skills and most call Sno Station programs; install sno-station-standalone@sno-ai instead if you only want the 7 skills that run without Sno Station. Approve nightly lessons with:

sno rem-reflect accept <id>

Skills declare what they need at the top of SKILL.md:

requires:
  programs: [reach]
  harness:
    - {slot: 4.shell, need: required}

Publishing a skill requires passing the repository checks (LICENSE/NOTICE present, no personal paths or hard-coded models, self-tests on a clean copy), after which the publish step writes PUBLISHED.json.

What are this agent's strengths and limitations?

How does this agent compare with similar options?

Key facts side by side with the most closely related agents.

How does FollowAgents rate this agent?

Recent releases

Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.

1. v1.0.31 Sno Station skills 1.0.31 2026-11-04
   Release notes excerpt

   Sno Station skills 1.0.31 The team skills your agents share: peer review, handoff, delegation, reporting and learning from mistakes. Install them with sno setup. Already installed? Run sno update.

   View on GitHub ↗
2. v1.0.30 Sno Station skills 1.0.30 2026-11-04
   Release notes excerpt

   Sno Station skills 1.0.30 The team skills your agents share: peer review, handoff, delegation, reporting and learning from mistakes. Install them with sno setup. Already installed? Run sno update.

   View on GitHub ↗
3. v1.0.29 Sno Station skills 1.0.29 2026-11-04
   Release notes excerpt

   Sno Station skills 1.0.29 The team skills your agents share: peer review, handoff, delegation, reporting and learning from mistakes. Install them with sno setup. Already installed? Run sno update.

   View on GitHub ↗