Claude Mods

*Doctrine you can execute. Knowledge that knows when it's stale. Agents that land their own work.*

claude-mods is a Claude Code plugin built on one premise: a skill that only *tells* the model something is a suggestion, and a skill that ships a gate is a guarantee. Its 112 skills carry the expert depth you'd expect - React, Rust, PostgreSQL, Cloudflare Workers, Next.js - but the ones that earn the install are the ones that run *agents*: a headless-worker spawner, a test-gated landing queue for parallel branches, a graduated-autonomy ladder for scheduled loops, and a security layer pointed at the agent itself. 63 skills ship executable tooling, 58 ship their own test suites, and 29 carry staleness verifiers that trip when the world moves under them.

Built on the Agent Skills specification (an open standard backed by Anthropic, Vercel, Google, Microsoft, and 40+ agent platforms) and Anthropic's patterns for long-running agents, it installs as a standard plugin and toggles off as easily as it toggles on.

3 agents. 112 skills. 13 styles. 15 hooks. 15 rules. One install.

Recent Updates

v3.9.0 (October 2026)
- 🐳 Docker family, rebuilt - docker-ops now covers Docker on your machine as well as in your repo: a Dockerfile and Compose auditor that carries hadolint, Trivy and Checkov IDs, a read-only doctor for Docker Desktop, OrbStack, Colima, Podman and Rancher, and verifiers for image tags and fast-moving facts. container-orchestration is now kubernetes-ops (Gateway API, Helm 4), and windows-ops gained a WSL reference.
- 🐘 ddev-ops skill - DDEV fails quietly: a default that moved under an unpinned project, a key it ignores, one developer's setting committed for the whole team. audit-ddev-config.py runs 26 checks at file:line, from the commonest in a read of 36 agency…