Gatekeeper for Sentry — error triage as a canonical agent workflow

[sergical]

I built a Sentry gatekeeper for Cloudflare OS and I'd like to gauge whether there's appetite for it upstream — or whether it's a useful test case for the "gatekeepers maintained independently from OS instances" future the README mentions.

What it unlocks: "What's crashing in prod, and why?" becomes a chat question. Agents search issues, read stack traces, and queue triage (resolve / archive / assign) that users batch-approve later — issue triage is almost the ideal fit for the async-approval + simulation model: high volume, low individual stakes, tedious for humans, fully revertible.

Why it's low-friction: it's token-based (instance URL + user auth token, works for SaaS and self-hosted), so unlike most gatekeepers there's no OAuth app for deployment admins to provision. Two granularities: whole org, or a single project.

Working implementation (built with your write-gatekeeper skill, follows the linear/homeassistant patterns, lint/build/tests/golden green): sergical#1.

(Disclosure: I work at Sentry on the DevEx team, some of this text was ai generated)