ship plugins

by Houseofmvps

All-in-one builder toolkit. Ship production-ready SaaS with one plugin.

by Helio

An agent-routed harness for end-to-end software product development.

by Jason Matthew

Stop-hook gate against trail-off: detects unpushed commits at turn end and injects an agent-directed nudge to review and finish the branch.

by wasikarn

A Claude Code harness that composes mattpocock-skills instead of duplicating it. Seven computational deny/ask gates (irrecoverable commands, subagent git guard, subagent spawn guard, task-complete separation, test integrity, config-write guard, codex-setup guard) plus a SubagentStop verdict-check gate, a 4 KB METHODOLOGY injected at SessionStart, and memory-lint for the auto-memory store.

by ckandrinirina

Spec-driven workflow plugin: design -> plan -> build -> ship. Transforms specs into architecture docs, epics/stories, and auto-generated expert and language-guide skills tailored to your tech stack. Story state lives in frontmatter; indexes are generated.

by kouko

Optional using-loom-code discovery routing and five stations carry one change from plan to PR with non-blocking cross-vendor suggestions, content-bound verification, one closing review, and a fast publication gate. Claude Code, Codex + Antigravity CLI.

by Chris Messina

Stage-keyed skills for the full Raycast-extension lifecycle: api-changelog (derive @raycast/api release notes Raycast doesn't publish), scaffold (new), develop (change code, modernize, house-style audit fix), ship (dep hygiene, house-style audit, Store compliance, PR, mirror, cleanup), greptile-loop (drive an open Store PR's automated review to 5/5), review-pr (inbound review of a third-party PR).

by dimonb

Drive one change end-to-end through a repo's pipeline — issue, spec, implementation, its own review passes, hand-off or merge — on GitHub or GitLab. Self-driven: every review pass runs as read-only subagents with adversarial verification, and nothing waits on an external reviewer.

by paulkim-lansik

Delivery-loop skills — land and release already-verified work through a repo's real gates. Tracker- and branch-model-agnostic: reads a consuming repo's own config instead of hardcoding one repo's setup.

by dimonb

Run the repo's ship skill in background terminals, using the same agent runtime as the parent session, with fleet monitoring, escalations, and automatic Codex parent continuity after transient model-capacity stops. Requires the ship skill; it launches nothing else.

by rajool

A toolkit of reusable Claude Code skills and agents: an invoke-only git workflow with guard hooks (no bulk add, no edits on main, no binaries/secrets), deep web research via the Gemini CLI, an Agent Skills authoring/validation kit, a project-agnostic file organizer, a macOS audio meeting recorder plus a processor that transcribes (ElevenLabs Scribe v2, with an automatic Gemini fallback for blocked networks), summarizes, and routes action items into the project's own tools, a document design system (typeset, formerly md-to-pdf) that gives every project its own house style and prints to PDF (a per-project theme of identity, logo, one accent, type pairing, mood and weight set — emphasis is a weight step, never bold — with cover page, letterhead or 16:9 deck slides, running head and folio, a table register with right-aligned tabular figures and ruled totals, financial statements from rows or a QuickBooks export, and first-class RTL/Persian typography — mirrored layout, x-height-matched pairs, a Latin guest face inside Persian sentences, Persian digits, Jalali dates, localized stationery labels), a claim-sourcing helper that links every factual claim to its original independent source, an end-of-day sweep that reconciles every repo with its remote, prunes stale branches, worktrees, and orphaned worktree directories (squash-merge aware, PR-aware), and brings every installed plugin up to its latest published version while flagging manifest/install version drift, a personal end-of-day retro-log engine (five questions, one dated file per day, built for pattern analysis over time), a Chrome browser-automation skill that drives Chrome via the DevTools Protocol (CDP) — snapshot/uid targeting, far more reliable and faster than coordinate-clicking extensions, with named persistent login profiles (isolated throwaway by default, user-defined per-account identities whose sessions survive across runs, guided new-machine onboarding, pinned server version, telemetry off) — a one-command installer that teaches a machine's global CLAUDE.md to render Persian/RTL chat replies correctly (whole-reply RTL widget card, atomic LTR token isolation, English-only plain chat text, BiDi-safe CLI fallback), and an always-on permission guard that blocks destructive commands like rm -rf, and a ship guard that refuses to end a turn while a requested "ship" is still unmerged. Skills act only when invoked.