Clownfish

Why it's ranked here

Its value comes from a careful split between judgment and authority. Codex proposes actions without GitHub write credentials, then deterministic checks revalidate live state before mutations. That design suits risky backlog maintenance, though the operational machinery is substantial and deliberately narrow.

What's good

The proposal-first workflow keeps uncertain items open and sends ambiguous cases to maintainers. Duplicate work guards, idempotent comments, stale-state checks, contributor credit, explicit merge gates, and separate security routing make automated cleanup traceable. Repair loops can update an existing branch or create a credited replacement.

Tradeoffs

Clownfish does not discover backlog clusters itself, and security-sensitive work must use another process. Effective operation depends on GitHub Actions, Codex, GitHub credentials, committed job definitions, environment gates, and careful promotion from planning to execution. Missing hydration, conflicts, failed checks, unclear intent, or stale state commonly stop automation for human review.

How to use it well

Use it as a second-stage maintainer system after humans or another tool group related work. Start in read-only planning, inspect the structured artifacts, then briefly open execution gates for approved clusters. It fits OpenClaw-specific issue cleanup and pull-request repair, not general backlog discovery or central security response.

Observed

Primary language

JavaScript using ECMAScript modules

Packaging

Private npm package requiring Node.js 24 or newer

Interface

Command-line workflows exposed through npm scripts

Automation platform

GitHub Actions coordinates planning, execution, repair, and finalization

Worker boundary

Codex proposes structured decisions without GitHub write credentials

Scope

Consumes preselected GitHub issue and pull-request clusters; does not discover the backlog itself

Testing

Package test command runs Node test files matching test/*.test.mjs