Team chat, from Slack, Linear and Sentry to GitHub

Work from Slack mentions and Linear issues; Sentry errors become issues.

The workflow of Team chat, from Slack, Linear and Sentry to GitHub: the events that start it, its agents and what they write.
The workflow of Team chat, from Slack, Linear and Sentry to GitHub: the events that start it, its agents and what they write.

The factory works where the team already talks. Someone mentions it in a Slack channel: the request becomes a GitHub issue, with its link in the thread. Someone assigns a Linear issue to openfabriq, like a teammate: an agent works in a fresh checkout of the repository, shows its progress in the issue's agent session, and answers there with the link of the pull request. A new production error in Sentry becomes a GitHub issue that explains its cause. A human reviews and merges on GitHub.

Slack, Linear and Sentry reach openfabriq through your relay, a Cloudflare Worker you deploy once with openfabriq connect relay (a free account is enough), or the same code on a server of yours. It keeps what comes in while your machine sleeps; the agents still run on your machine.

What this factory does

The GitHub issues, from Slack or from Sentry, then enter your usual flow: combined with 02-sdlc (see below), issue.opened starts triage, and the implementation starts once the issue is ready-for-agent. A request made in Slack is checked before any code is written, and leaves a trace on GitHub. The engine opens the issue with the gh account.

Least privilege: the Slack automation may only reply and open an issue, and its agent only reads; the Linear one may only answer and open a pull request; the Sentry one may only open an issue and add a note. None can push to an existing pull request or review one: validate refuses it for runs that come from Slack, Linear or Sentry. The investigator has no edit tool. Before the implementer's pull request leaves, the engine runs checks itself: a failing check goes back to the agent, and if it still fails, no pull request is opened.

Stack traces can carry personal data (emails, ids in URLs). The engine leaves out frame variables, and the investigator is told not to copy personal data into the issue; read the issues it opens with that in mind.

Tree

openfabriq/
├── factory.yaml                              integrations (slack, linear, sentry), setup, commands, checks
├── agents/
│   ├── intake/agent.md                       turns a Slack request into a GitHub issue
│   ├── implementer/agent.md                  delivers a clear Linear request, asks when it is not
│   └── investigator/agent.md                 finds the cause of an error, fixes nothing
├── automations/
│   ├── answer-slack-mentions/                when: slack.mentioned
│   ├── implement-delegated-linear-issues/    when: linear.issue.delegated
│   ├── investigate-sentry-errors/            when: sentry.issue.created level:error environment:production
│   ├── rework-pull-requests/                 when: pull_request.changes_requested
│   └── resolve-conflicts/                    when: pull_request.conflicted
└── skills/
    └── test-driven-development/SKILL.md      test first, red then green, the failing run as proof

Who can start it

maintainers are GitHub logins; they do not apply to Slack, Linear and Sentry. There, the gate is the place: anyone in the Slack channel, anyone of the Linear team, any error that reaches the Sentry project. Use a private channel. openfabriq plan lists adding or changing slack, linear and sentry as security changes.

Make it yours

  1. Copy the openfabriq/ folder to the root of your repository.
  2. In factory.yaml, replace repository, maintainers, the Slack channel, the Linear team key, the Sentry organization and project (add url for a self-hosted Sentry), the setup command and the commands. Adjust the Sentry query: it is Sentry's own search, times_seen:>50 works too.
  3. Run openfabriq connect relay, then openfabriq connect slack, openfabriq connect linear and openfabriq connect sentry: they guide you, test the tokens, give the relay what it keeps and save the rest in ~/.openfabriq/secrets.json. Installing openfabriq in Linear needs an admin of the workspace.
  4. In Linear, Settings, Integrations, GitHub, add the repository: each issue then lists its pull request.
  5. Invite the bot to the channel: /invite @openfabriq.
  6. Run openfabriq plan, openfabriq apply, then openfabriq start.

With 02-sdlc

A repository has one openfabriq/ folder. To add the team chat to the 02-sdlc factory, copy into it:

  • under integrations in its factory.yaml, the slack and linear keys (it already has sentry);
  • agents/intake/ and automations/answer-slack-mentions/;
  • agents/implementer/ as agents/linear-implementer/, since 02-sdlc has its own implementer, for the issues triage found ready; and automations/implement-delegated-linear-issues/, with agent: linear-implementer.

Leave out this factory's investigator and investigate-sentry-errors: 02-sdlc has its own, investigate-production-errors. Then run openfabriq validate and openfabriq plan.

More in Getting started

All templates
The workflow of At a glance
00

At a glance

The whole format in a few files: issues and production errors become pull requests, failed deploys become issues.

issue.labeledSentrya webhook

3 agents3 automations

openfabriq init --example 00-at-a-glance
The workflow of Quickstart
01

Quickstart

A maintainer labels an issue, an agent opens a pull request and changes it on a comment that mentions @openfabriq.

issue.labeledpull_request.commented

1 agent3 automations

openfabriq init --example 01-quickstart
The workflow of SDLC, from issue to ready pull request
02

SDLC, from issue to ready pull request

The full flow: triage, investigation, implementation reviewed adversarially before the pull request exists, conflicts resolved, with checks and screenshots as proof.

Issuespull requestsSentry

4 agents7 automations

openfabriq init --example 02-sdlc