GitHub Projects Beat Linear When Agents Need Hooks

The interesting part of this Skill Recordings setup isn’t the GitHub Project board. It’s that the board is becoming part of the agent loop.

Joel was working in skillrecordings/support with clawdbot, later moltbot, and syncing work into a GitHub Projects board. GitHub webhooks feed changes back into the bot so it can keep monitoring the situation instead of waiting for a human to remember what moved.

That’s the clever bit: GitHub CLI and GitHub webhooks are boring, scriptable, and already sitting next to the code. Linear might be nicer as a dedicated tracker, but this keeps the work close to GitHub, easy for Claude to inspect, and easy for Joel to groom without another integration tax.

For joelclaw, this maps cleanly to the same pattern as agent-run review loops: the tracker is not just a dashboard, it’s an event source. If the board emits state changes and the bot can react, the project plan becomes operational instead of decorative.

Key Ideas

• GitHub Projects can act as an agent-readable work queue when paired with GitHub CLI and GitHub webhooks.
• Keeping skillrecordings/support and its project board inside GitHub reduces the integration surface compared with moving issue flow into Linear.
• Feeding project updates back into clawdbot / moltbot turns issue movement into a monitoring signal instead of passive project-management noise.
• The pattern fits joelclaw agent-loop work because the same state changes humans use for grooming can drive bot review, follow-up, and status checks.

Links

• Skill Recordings GitHub Project #3
• skillrecordings/support
• GitHub Projects documentation
• GitHub CLI
• GitHub webhooks documentation
• Linear
• Claude
• joelclaw