Epic: Starter ops/DevOps agents — one-click Trinity instance management onboarding

[vybe]

Summary

Ship a curated set of ready-to-deploy ops / DevOps starter agents so a new operator can stand up an agent that manages their own Trinity instance — health, updates, logs, rollback, provisioning — in minutes, on whatever platform they run Trinity on. This builds directly on the existing public ops agent (abilityai/trinity-ops-public) and the trinity plugin scaffolding in abilityai/abilities (/trinity:deploy-new-instance, /trinity:onboard), turning ad-hoc scaffolding into a first-class, one-click onboarding experience.

Context

Trinity is sovereign, per-deployment infrastructure: every operator runs their own instance. The hardest part of adoption is the first hour — connecting, deploying, and learning to operate the platform. We already proved the model with the Claude Code ops agent (abilityai/trinity-ops-public) and the /trinity:deploy-new-instance skill, which scaffolds an ops agent to manage an instance. This epic packages that into a set of starter agents an operator can launch with minimal setup, across the platforms they actually deploy on, so "manage your Trinity" is a one-click starting point rather than a from-scratch build.

Scope confirmed with requester (2026-06-15): focus the starter set on the instance ops / DevOps archetype (not role agents, channel-demo chat agents, or a general gallery), in the public tracker for community contribution.

Children (sub-issues)

• Package the ops/DevOps agent as a first-class one-click starter
• Cross-platform bootstrap for the ops starter (Linux server / macOS / cloud VM & Docker host)
• "Operate your Trinity in 5 minutes" onboarding quickstart + sample ops prompts

Acceptance Criteria

• A new operator can deploy a working instance-management agent from a documented, near-one-click flow
• The starter works across the common platforms operators run Trinity on (self-hosted Linux, macOS, cloud VM / Docker host)
• Onboarding docs let a first-time user go from zero to a running ops agent in ~5 minutes
• Starter agent capabilities cover the core ops loop: health checks, updates, log inspection, rollback, basic provisioning
• Reuses and consolidates the existing trinity-ops-public agent and /trinity:deploy-new-instance / /trinity:onboard scaffolding rather than duplicating them

Technical Notes

• Existing assets to build on (all public): abilityai/trinity-ops-public (the ops agent), abilityai/abilities → trinity plugin (/trinity:connect, /trinity:onboard, /trinity:deploy-new-instance, which already "scaffold a complete ops agent to manage it").
• Local starter templates live under config/agent-templates/; the platform already supports template-based agent creation (POST /api/agents with a template, including github:Org/repo).
• Decide whether the ops starter ships as a Trinity template (in config/agent-templates/), a pinned github: template repo, or both.
• Keep agent-defined pipeline / heartbeat work owned by the agent (CLAUDE.md Rule security: implement safe tar extraction with symlink/hardlink validation #8) — Trinity provides the read surface, not an orchestrator.