triagelisted

# Triage — Production Errors → Linear → Worktree → PR Production monitoring (Sentry/Datadog) feeds into agents that triage and propose fixes. You already have dev-time observability (cost tracking, trace coverage, ci-minutes audits). This skill adds **runtime** observability — what's actually happening in users' browsers and your servers. ## Prerequisites (one-time setup) **Check these are actually available before doing anything.** This skill needs the Sentry MCP server (for errors) and, to file tickets, the Linear MCP server. If the `mcp__sentry__*` tools are missing, say so and stop: > `/bs:triage` needs the Sentry MCP server, which isn't configured in this session. > Set it up per the instructions below, then re-run. Do not fabricate errors or file tickets against a service you cannot reach. ### 1. Sentry MCP server ```bash # Claude Code — Sentry's hosted MCP at mcp.sentry.dev (HTTP transport) claude mcp add --transport http sentry https://mcp.sentry.dev/mcp # Verify /mcp # Should show: sentry — connected ``` Auth: Sentry MCP uses OAuth — the first tool call triggers a browser flow to grant access to your org. Scoped tokens are stored in `~/.claude/.mcp-auth/`. If Sentry MCP isn't yet stable or you prefer self-hosted: install `@sentry/mcp-server` from npm (Sentry maintains an official package) — fall back to API-token mode: ```bash claude mcp add --transport stdio sentry \ --env SENTRY_AUTH_TOKEN=$SENTRY_AUTH_TOKEN \ --env SENTRY_ORG=$SENTRY_ORG \ -- npx