openclaw/clawsweeper

ClawSweeper scans all issues and PRs and suggest what we can close, and why. It runs every PR / Issue once a week.

What it solves

ClawSweeper is a conservative maintenance bot designed to reduce the manual burden on maintainers of OpenClaw repositories. It automates the review of open issues and pull requests, tidies GitHub comments, and handles the repair and merging of trusted code changes while ensuring that no action is taken without strict guardrails.

How it works

The system operates through a combination of scheduled scans and event-driven triggers. It uses a model (Codex) to review issues and PRs, generating durable markdown reports and syncing a single, edited-in-place public comment to avoid clutter.

Key operational flows include:

  • Review & Apply: It analyzes items and proposes actions (like closing duplicates or stale items). It only executes these actions after re-verifying the live state against repository-specific profiles.
  • Repair & Automerge: Maintainers can opt PRs into autofix or automerge. The bot then enters a bounded loop of review and repair, validating the code before finally merging it once all policy gates and CI checks are green.
  • Issue-to-PR: For high-confidence bug reviews, the bot can automatically open guarded implementation PRs, establishing a failing regression before proposing a fix.
  • Maintainer Steering: Maintainers control the bot via GitHub comments (e.g., @clawsweeper fix, @clawsweeper automerge), which are routed through a Cloudflare Worker to GitHub Actions executors.

Who it’s for

It is primarily for maintainers of the OpenClaw ecosystem and other developers who wish to self-host a conservative, AI-driven maintenance bot for their own GitHub organizations.

Highlights

  • Conservative Guardrails: Prevents automatic closing or merging of maintainer-authored items and requires strict evidence for close proposals.
  • Marker-Backed Comments: Edits a single comment in place using hidden markers to track state, preventing comment spam.
  • Bounded Repair Loop: Uses a Codex-driven review/fix loop for PR repairs that must pass deterministic executor steps before mutation.
  • Automated Issue Implementation: Can automatically transform verified bugs into implementation PRs with regression tests.
  • Observability Dashboard: Provides a detailed view of system flow, worker capacity, and pipeline status via a Cloudflare-hosted dashboard.

Related

  • Project
  • Project
  • Project
  • Project