latitude-setup
Latitude Setup (zero-account onboarding)
Outcome
You are done when the user has, in one final message from you:
- Telemetry installed in their app or agent harness and verified with the Latitude CLI against a real run.
- A first Artifact:
artifacts/first-session.html, built from the bundled template with the data of the clean session you just verified, so the user sees what Latitude captured (model calls, tool calls, tokens, cost, timing, the conversation) without opening anything else. When the account is temporary, the page carries the Claim your workspace button. - The claim link and its expiry (temporary-account flow), and any restart the user still owes.
Installing telemetry is the prerequisite. The Artifact is how the user sees the value in the first five minutes; it is part of the deliverable, built by default, and skipped only when the user explicitly says so. Verifying traces and building the Artifact are two separate steps (8 and 9): verification proves the pipeline with the CLI, the Artifact shows the user what came through.
This skill orchestrates the from-scratch path: the user has no Latitude account and no API key. It provisions a temporary account via the CLI, instruments the target, verifies real traces, cleans up, builds the Artifact, and returns a claim link.
The target is whatever should emit traces: an app (instrumented with the Latitude SDKs or an OTLP exporter) or an agent harness the user runs locally (Claude Code, Hermes, OpenClaw, Pi, Prime Intellect, each with its own plugin). Everything below applies to both; steps call out where a harness differs.
This skill depends on latitude-cli (install + auth + command primitives) and latitude-telemetry (instrumentation). Read both; this skill adds the orchestration between them. The first Artifact needs nothing beyond this skill's own first-artifact.html template and the CLI. For richer or refreshable reports later, the separate latitude-artifacts skill exists; it is not needed here.