agent-plugins is an open-source (MIT) plugin marketplace from Stylus Nexus: ten plugins and 62 skills that install natively into Claude Code and OpenAI Codex, and into about seventy other coding agents through the Skills CLI. Eight are skill packs for building, shipping, deciding what to build, and getting it noticed; two are tools.
Eight packs, grouped by the job rather than the technology. Each pack's README carries a rule-ownership table, so the skills inside it don't compete for the same trigger. Dashed chips are agents, which run in Claude Code.
The daily loop: read the issue, ground assumptions in the live database, prove it works, then review, merge, and promote. One issue or a batch.
The last mile: turning agent output into something a person wants to read, and getting their judgment back.
Built on one premise: a single model's confident answer is not evidence. Councils, arbitration, and specs checked against the actual codebase.
The unglamorous pre-launch gates: a missing rate limit, an unsigned webhook, a compromised transitive dependency.
Deciding the version, waiting on CI, publishing, and keeping dependencies current between releases.
Building an accurate picture of a codebase, and the libraries it leans on, before committing to a change.
Deciding what to build and why: vision, strategy, value, objectives, roadmaps, and discovery. A chief PM that teaches as it goes and labels evidence apart from assumption.
Getting a product noticed on a small team's hours: positioning, copy, launches, community, outreach, email, and measurement. It drafts; it never posts or sends for you.
Two plugins ship executable code rather than markdown, so they live in their own repositories and install from there.
Track-aware daily planning over GitHub issues: shared tracks, AI clustering, coverage, and dependency-aware next-up. Pure-stdlib Python CLI plus an accessible VS Code viewer with a repo-qualified dependency graph.
Language-aware defect hunter. Detects the stack, triages files by risk, runs the real analyzers (ruff, mypy, tsc, eslint, rubocop, brakeman), then reports findings in confidence tiers, correlated against existing issues.
Add the marketplace once, then take whichever packs you want. Plugins run code on your machine, so review a plugin's source before installing it.
Terminal, VS Code, and JetBrains extensions share one config, so one install covers all three.
/plugin marketplace add stylusnexus/agent-plugins /plugin install ship-pipeline@stylus-nexus
Skills carry the pack name: /ship-pipeline:prove-it. To get updates automatically, open /plugin → Marketplaces → stylus-nexus → Enable auto-update.
CLI, app, and IDE extension. Codex reads its own index, so the same marketplace works unchanged.
codex plugin marketplace add stylusnexus/agent-plugins codex plugin add ship-pipeline@stylus-nexus
Type $ship-pipeline:prove-it or run /skills. In the ChatGPT app, type @. To update, run codex plugin marketplace upgrade stylus-nexus, then the add command again.
Cursor, Copilot, Gemini CLI, Windsurf, Zed, opencode, Cline, Continue, Hermes, Goose, Warp, Aider and more. Detects which agents you already have.
npx skills add stylusnexus/agent-plugins npx skills add stylusnexus/agent-plugins --skill prove-it deploy
Skills arrive without the pack name: /prove-it. Update with npx skills update. Covers all eight skill packs.
Pick the goal closest to yours. If you only take one pack, take ship-pipeline.
Ship code every day without breaking things
ship-pipelinesecond-opinion
Get ready to launch
hardeningrelease-opsgo-to-market
Decide what to build, and why
product-strategysecond-opinion
Get a product noticed on a few hours a week
go-to-marketreporting-comms
Find your way around an unfamiliar codebase
codebase-intelreporting-comms
Make agent output easier to read and review
reporting-comms
They're repo-agnostic: every skill detects your conventions rather than assuming them, and defers to a repo-local version of itself when your project defines one. They still hold a point of view: evidence before "done", detect rather than assume, one rule per owner. Each README states the opinion plainly, so you can tell quickly whether it fits how you work.
Run /plugin marketplace add stylusnexus/agent-plugins, then /plugin install ship-pipeline@stylus-nexus for each pack you want. Skills then carry the pack name, as in /ship-pipeline:prove-it.
Run codex plugin marketplace add stylusnexus/agent-plugins, then codex plugin add ship-pipeline@stylus-nexus for each pack you want. Invoke a skill with $ship-pipeline:prove-it, or run /skills.
Run npx skills add stylusnexus/agent-plugins. The Skills CLI detects which agents you have and writes the skills into each one; add --skill and skill names to take only some.
In Claude Code, enable auto-update for the stylus-nexus marketplace under /plugin, Marketplaces. In Codex, run codex plugin marketplace upgrade stylus-nexus, then add each pack again. With the Skills CLI, run npx skills update.
ship-pipeline is the agent-plugins skill pack for the daily loop: it reads the issue, grounds assumptions in the live database, proves the change works, then reviews, merges, and promotes it, for one issue or a batch.
Claude Code and Codex plugins prefix skills with the pack name. Skills installed through the Skills CLI or npm have no prefix.