Shared playbooks and outputs for the whole workspace
Every infrastructure team has prompts that get rewritten over and over: how to triage a failed plan, how to write a postmortem, how to summarize drift, how to explain a risky change before merge. Workspace Skills are now a first-class way to turn those prompts into shared Atmos playbooks for your team.
Create a skill once in workspace settings and it becomes available anywhere your team uses Atmos AI: in the dashboard chat and through the Atmos Pro MCP server installed in Claude Code, Codex, Cursor, Continue, Cline, or another MCP client.
Skills are workspace-scoped, not repository-scoped. That means a platform team can define one
/triage or /postmortem skill and use it across every repository, stack, component, and environment in the workspace.Skills also pair with shared artifacts. When an agent produces a durable output such as a postmortem document, incident summary, rollout note, or investigation handoff, it can save that artifact back to Atmos Pro so the team can find it later instead of losing it in one person's chat history.
Use cases
- Triage. Standardize how agents inspect failing workflow runs, drifted instances, dispatch failures, pending approvals, and permission problems before they answer.
- Root cause analysis and postmortems. Give the team a shared incident-writing format so every RCA leads with customer impact, separates symptoms from causes, and closes with concrete follow-up.
- Drift summaries. Turn a noisy list of changed resources into a concise explanation of what changed, why it matters, and what should happen next.
- Deployment readiness. Ask an agent to review recent runs, affected stacks, approvals, and open pull requests before a production rollout.
- Pull request review. Keep review output consistent across repositories by encoding what your team cares about: Atmos stack shape, component boundaries, workflow behavior, and operational risk.
- Handoffs. Convert investigation context into a clear status update for the next engineer, with links back to the relevant runs, instances, and repositories.
- Shareable artifacts. Save the final output of a skilled investigation as a team-visible artifact: a PDF postmortem, an incident brief, a deployment readiness note, or a reusable investigation summary.
Available from the MCP too
Skills are not just an in-product chat shortcut. Anyone who installs the Atmos Pro MCP for the workspace can list and use the same workspace skills from their agentic editor.
That makes Skills a repository-agnostic way to distribute team knowledge. Instead of copying prompt snippets into every repo's docs or asking each engineer to maintain their own local prompt library, the workspace becomes the source of truth.
Artifacts make the work durable
Chats are good for investigation, but the result often needs to live somewhere teammates can reuse it. Artifacts turn the output of a skill into a workspace object: something attached to the same operational context, attributable to the skill that produced it, and shareable with the team.
For example,
/postmortem can guide the analysis, then save the finished RCA as a PDF artifact. A triage skill can produce a handoff note for the on-call rotation. A deployment-readiness skill can save the checklist the release owner used before promoting a change.The important part is that the output is no longer trapped in one chat session or one repository. It belongs to the workspace.
Built-ins, customization, and control
Atmos Pro ships built-in system skills, starting with
/postmortem. Workspace admins can clone a system skill to customize the voice, structure, and output for their team. A workspace skill with the same slug shadows the system skill, so /postmortem can mean "our postmortem format" without changing how engineers invoke it.Admins can also enable or disable visible skills. Disabled skills stay in settings for editing and history, but they disappear from the slash-command menu and are unavailable through MCP until re-enabled.
Skills are managed in Settings -> AI -> Skills.
