@orgloop/module-engineering

Engineering organization workflow module for OrgLoop. Wires GitHub, Linear, and Claude Code into a supervised loop through an OpenClaw agent — PR reviews get addressed, CI failures get fixed, tickets get triaged, and Claude Code sessions get evaluated and relaunched when they finish.

5 routes, 4 SOPs, and 3 transforms. Install it, configure your connectors, and the engineering loop runs.

Install

orgloop add module engineering
orgloop doctor    # shows what's needed and how to get it
orgloop start

What You Get

Routes

| Route | Trigger | What Happens | |-------|---------|-------------| | pr-review | PR review submitted | Agent wakes with PR review SOP — reads comments, makes fixes, pushes, re-requests review | | pr-comment | Inline PR comment | Same SOP — addresses the specific comment in context | | ci-failure | CI workflow fails | Agent wakes with CI failure SOP — diagnoses the failure, fixes root cause, pushes | | linear-triage | Linear issue updated | Agent wakes with triage SOP — evaluates the ticket, decides action based on state change | | claude-code-supervisor | Claude Code session ends | Agent evaluates what the dev session accomplished — relaunches, opens PR, or escalates |

The Loop in Action

This is where it clicks. The routes aren't independent — they form a loop:

Linear ticket assigned
  → linear-triage route wakes agent → agent starts Claude Code implementation
    → Claude Code finishes → claude-code-supervisor route wakes agent
      → agent evaluates output → opens PR
        → PR gets review → pr-review route wakes agent
          → agent addresses feedback → pushes → CI runs
            → CI fails → ci-failure route wakes agent
              → agent fixes → CI passes → ready for merge

Each step triggers the next. The org loops.

SOPs (Standard Operating Procedures)

Each route delivers a focused SOP — the agent gets specific instructions for exactly what happened, not a grab-bag of everything it might need to know.

• pr-review.md — Parse the review, address each comment individually, check PR health before re-requesting review. Includes guidance on AI/bot comments and security red flags.
• ci-failure.md — Categorize the failure (type errors, lint, tests, build, flaky), fix root cause, verify CI passes. Never skip or disable tests.
• linear-ticket.md — Triage before acting: is it clear? Is it blocked? Is someone already on it? Action table by state change.
• claude-code-supervisor.md — Evaluate session output, decide whether to relaunch, open PR, or escalate. Guards against infinite relaunch loops.

Transforms

• drop-bot-noise — Filters out events authored by bots (you don't need to review your own bot's actions)
• my-notifications — Keeps only events involving your watched GitHub accounts (GITHUB_WATCHED)
• dedup — Deduplicates events within a 10-minute window (GitHub often sends multiple events for the same action)

Prerequisites

• GitHub — repo access + personal access token with repo scope
• OpenClaw — running locally with a webhook token configured
• Linear (optional) — API key for ticket management. Module works without it; Linear routes are skipped.
• Claude Code (optional) — installed locally for the supervisor loop. Module works without it; supervisor route is skipped.

Run orgloop doctor after installing — it checks every dependency and tells you exactly what's missing and how to get it.

Parameters

| Parameter | Required | Default | Description | |-----------|----------|---------|-------------| | github_source | yes | github | Name of your GitHub source connector | | linear_source | no | linear | Name of your Linear source connector | | claude_code_source | no | claude-code | Name of your Claude Code source connector | | agent_actor | yes | openclaw-engineering-agent | Name of your agent actor |

Required Credentials

| Variable | Required | Description | |----------|----------|-------------| | GITHUB_TOKEN | yes | GitHub personal access token (repo scope) | | GITHUB_REPO | yes | GitHub repository (owner/repo) | | GITHUB_WATCHED | yes | Comma-separated GitHub usernames to watch | | OPENCLAW_WEBHOOK_TOKEN | yes | OpenClaw webhook authentication token | | OPENCLAW_AGENT_ID | yes | OpenClaw agent ID | | LINEAR_API_KEY | no | Linear API key | | LINEAR_TEAM_KEY | no | Linear team key |

Customizing

The SOPs are starting points. Copy them to your project's sops/ directory and customize:

cp node_modules/@orgloop/module-engineering/sops/*.md ./sops/

Then update your routes to point at your local copies. Make the SOPs match how your team actually works — add your review standards, your CI conventions, your triage criteria.

Documentation

Full documentation at orgloop.ai. See the Manifesto for the thinking behind Organization as Code.

License

MIT