Plug'n Skills
Ready-to-install skills and plugins that make coding agents better at real
development work.
Plug'n Skills is a library of plugin packs for Codex, Claude Code, Cursor, and
other coding agents. The source tree is agent-agnostic: every pack works from
any of those hosts and is never locked to one of them. Each pack gives an
agent a focused workflow: what to inspect, which
commands to run, what to verify, and when to use a deterministic helper instead
of improvising from a prompt.
Use it when you want an agent to handle more than generic code edits:
- build, debug, profile, test, package, and release Swift apps;
- work with Kotlin Multiplatform, Tauri, PixiJS, Codex CLI, and Claude Code;
- review architecture, product direction, interface quality, and game systems;
- plan scientific research, spec-driven delivery, context compression, and
agent capability synthesis.
The repository ships 16 installable plugin packs and 150+ focused agent
skills, all plain repository content: manifests, SKILL.md files, references,
validators, and helper scripts. Inspect it, validate it from a fresh clone,
install only the packs you need, and keep generated local marketplace or cache
state out of the repo.
Quick Start
Easiest: Ask Your Agent
Paste this prompt into the coding agent you already use:
Install Plug'n Skills from https://github.com/Xopoko/plug-n-skills on this
computer. Follow the repository instructions for the coding agent you are
running in, validate the source tree first, use a dry run before writing global
plugin state, install only the plugin packs I request unless I ask for all of
them, and report exactly what was changed.
Clone And Validate
git clone https://github.com/Xopoko/plug-n-skills.git
cd plug-n-skills
python3 scripts/validate-repository.py
Codex
Preview the Codex install plan before writing global state:
python3 scripts/install-codex-plugins.py --dry-run
Install or refresh every plugin from this checkout:
python3 scripts/install-codex-plugins.py
python3 scripts/install-codex-plugins.py --check-only
Install only selected plugin packs:
python3 scripts/install-codex-plugins.py --plugin capability-workbench
python3 scripts/install-codex-plugins.py \
--plugin kotlin-multiplatform \
--plugin spec-driven-development
Exclude plugin packs that are not useful on the current host:
python3 scripts/install-codex-plugins.py \
--exclude-plugin build-swift-apps \
--exclude-plugin kotlin-multiplatform
The installer validates the repository, generates a local Codex marketplace
file at .agents/plugins/marketplace.json, points Codex's local marketplace
at this checkout, enables the selected plugins, and materializes cache entries
under the active Codex home: $CODEX_HOME when it is nonempty, otherwise
~/.codex. A configured CODEX_HOME must resolve to an existing directory.
Explicit --config-path and --cache-root values take precedence.
.agents/ and Codex cache directories are local runtime state, not part of
the published source tree.
Claude Code
Add this repository as a Claude Code marketplace, then install the pack you
need:
/plugin marketplace add Xopoko/plug-n-skills
/plugin install capability-workbench@xopoko-plug-n-skills
/reload-plugins
For a local checkout, use its path instead of the GitHub shorthand:
/plugin marketplace add /path/to/plug-n-skills
Claude Code reads the root .claude-plugin/marketplace.json and each plugin's
.claude-plugin/plugin.json.
Cursor
Cursor consumes SKILL.md folders directly and has no plugin marketplace.
Install (or refresh) the repository skills into Cursor's global skills
directory:
python3 scripts/install-cursor-skills.py --dry-run
python3 scripts/install-cursor-skills.py
python3 scripts/install-cursor-skills.py --check-only
Use repeated --plugin or --exclude-plugin flags when a host should only see
part of the repository.
The installer is idempotent: unchanged skills are skipped, drifted skills are
replaced to match the repository source, and repeated runs converge.
Included Plugin Packs
| Plugin | Use it for |
|---|
architecture-intelligence | Source-backed architecture audits, async state consistency, ownership and runtime topology, module boundaries, ADRs, fitness functions, conformance checks, and refactoring strategy. |
build-swift-apps | Building, debugging, profiling, testing, packaging, and releasing Swift apps across iOS and macOS. |
capability-workbench | Capability discovery, synthesis, plugin packaging, agent guidance files, trigger metadata, Codex catalog-budget and exact evidence-coverage audits, install-scope decisions, vetting, repair, icon workflows, install/cache checks, and separate runtime-discovery checks. |
claude-code | Claude Code CLI operations, print-mode automation, diagnostics, plugin and MCP lifecycle, hooks, settings, agents, sessions, and worktrees. |
codex-cli | Codex CLI operations, automation, producer-native deferred completion, diagnostics, live thread supervision with protected-policy gates, version-aware skill handoffs with source/runtime separation, evidence corrections, aggregate claim boundaries, plugin and MCP lifecycle, normalized session trace audits, and local environment actions. |
scheduled-automation | Local scheduler diagnostics, real-runtime proof, safe canaries, correlated run receipts, missed-run analysis, and rollback-aware repair. |
gitlab-review | Race-safe GitLab merge request review response, reviewer-owned resolution, idempotent thread replies, and exact-head handoff proof. |
stacked-delivery | Dependency-safe stacked change delivery with exact parent-head binding, descendant invalidation, node-local proof, post-rewrite proof-record freshness, distinct provenance and publication-authority gates, safe landing order, and guarded handoff receipts. |
git-worktree-safety | Git pointer and SSH-signed-commit recovery with redacted evidence, exact-state journals, one-use authorization, and fail-closed proof. |
context-density | Context design, long-context placement, typed state and companion-drift validation, research-backed acceptance gates, prompt contracts, skill compression, structural handoff, and validation reporting. |
design-intelligence | Product framing, interface architecture, interaction design, visual hierarchy, accessibility, and design-system governance. |
See plugins/README.md for the per-plugin source index and
manifest identifiers.
Token Efficiency
This collection is designed around progressive disclosure. Agents can
route from lightweight metadata first, then load the selected
SKILL.md body only for the chosen workflow.
These estimates are generated with scripts/token-report.py using
tiktoken and the o200k_base encoding. Different agents may
wrap metadata differently, so the exact number is less important than
the split between always-visible routing metadata and on-demand skill
instructions.
| Metric | Count | Tokens | Notes |
|---|
| Plugin packs | 19 | - | Installable packages under plugins/. |
| Skill entrypoints | 171 | - | SKILL.md files exposed through plugin metadata. |
| Reference files | 241 | - | Longer ledgers, contracts, scorecards, and source notes. |
| Helper and validator scripts | 88 | - | Deterministic plugin-local helpers. |
| Startup metadata | 171 skills | 13,993 | Skill name, description, and file pointer for routing. |
| On-demand skill bodies | 171 skills | 124,665 | Instruction bodies after frontmatter, loaded only when selected. |
Regenerate the report after skill edits:
python3 scripts/token-report.py
Plugin Token Rollup
Descriptions are split from the numeric rollup so GitHub does not
compress long prose into narrow table cells.
Token columns are startup metadata / on-demand body.
| Plugin | Skills | Refs | Scripts | Startup | Body |
|---|
build-swift-apps | 61 | 90 | 36 | 4,743 | 36,116 |
pixijs | 26 | 64 | 0 | 1,861 | 7,967 |
tauri | 6 | 0 | 1 | 481 | 3,235 |
scientific-research | 1 | 4 | 1 | 91 | 2,024 |
context-density | 1 | 9 | 8 | 137 | 2,806 |
capability-workbench | 10 | 15 | 23 | 958 | 13,076 |
codex-cli | 8 | 5 | 3 | 688 | 10,059 |
scheduled-automation | 1 | 5 | 0 | 103 | 1,158 |
gitlab-review | 1 | 3 | 1 | 117 | 1,308 |
stacked-delivery | 1 | 4 | 1 | 123 | 2,100 |
git-worktree-safety | 2 |
Plugin Focus
| Plugin | Description |
|---|
build-swift-apps | Agent skills for building, debugging, profiling, testing, refactoring, and shipping Swift apps across Apple platforms. |
pixijs | PixiJS v8 skill collection: application setup, scene graph, assets, events, filters, rendering, performance, v7->v8 migration, and project scaffolding. |
tauri | Tauri 2 development: project setup/migration, tauri.conf & capabilities/security, Rust IPC & plugins, shell UI, debug/test, and distribution/mobile release. |
scientific-research | Disciplined scholarly research workflow: discovery, source routing, DOI dedup, claim ledgers, and evidence quality gates across arXiv/OpenAlex/Crossref/Europe PMC/Semantic Scholar/NCBI/CORE/OpenCitations. |
context-density | Context design, prompt-contract optimization, typed state validation, research-backed acceptance gates, compression validation, and structural handoff for agent skills, plugins, prompts, docs, and workflows. |
capability-workbench | Agent-agnostic capability workbench for discovery, synthesis, portfolio architecture, trigger metadata, Codex catalog-budget and evidence-coverage audits, agent guidance, vetting, repair, installation, packaging, and imagegen-backed plugin icons. |
codex-cli | Codex CLI operations, automation, producer-native deferred completion, diagnostics, live thread supervision with protected-policy gates and version-aware skill handoffs, plugin and MCP lifecycle, normalized session trace audits, and local environment actions. |
scheduled-automation | Local scheduler diagnostics, real-runtime proof, safe canaries, correlated run receipts, and missed-run analysis for launchd, systemd timers, cron, and Windows Task Scheduler. |
gitlab-review | Race-safe GitLab merge request review response with complete discussion inventory, reviewer-owned resolution, idempotent replies, and exact-head handoff proof. |
stacked-delivery | Dependency-safe delivery for stacked pull requests, merge requests, and dependent changes with exact parent-head binding, descendant invalidation, per-node proof, post-rewrite proof-record freshness, provenance and publication-authority boundaries, safe landing order, and guarded handoff receipts. |
git-worktree-safety | Fail-closed Git safety for worktree pointer and SSH-signed-commit recovery, with redacted evidence, exact-state journals, one-use authorization, and explicit host boundaries. |
Skill Token Index
Token cells are shown as startup/body.
build-swift-apps
| Skill | Tokens | Description |
|---|
app-icon-studio | 70/984 | Create, generate, evaluate, export, install, or debug iOS and macOS app icons, including AppIcon.appiconset assets and macOS .icns bundle icons. |
apple-dev-research | 67/503 | Search Apple Dev Search for Swift, SwiftUI, Xcode, iOS, macOS, and Apple-platform community articles, tutorials, blogs, and write-ups. |
apple-firmware-inspector | 87/676 | Apple firmware and binary reverse engineering with the ipsw CLI: IPSW/kernelcache download/extraction, dyld_shared_cache disassembly, private headers, entitlements, Mach-O analysis, Apple internals, KEXTs, and security research. |
appstore-ads-operator | 72/843 | Manage Apple Ads with asc ads: separate auth, org lookup, campaigns, ad groups, ads, keywords, reports, raw API requests, and safe live testing. |
appstore-archive-uploader | 73/800 | Manage Xcode version/build numbers, archive, export, upload, and publish IPA/PKG artifacts with asc xcode helpers before TestFlight or App Store submission. |
appstore-aso-auditor | 80/687 | Run an offline ASO audit on canonical App Store metadata under ./metadata and, when available, add Astro MCP keyword-gap analysis and Apple app-tag context. Use after asc metadata pull. |
appstore-build-monitor | 60/334 | Track build processing, find latest builds, and clean up old builds with asc. Use when managing build retention or waiting on processing. |
appstore-connect-cli | 65/521 | Use asc CLI for App Store Connect command discovery, auth, output formats, pagination, schemas, canonical verbs, Apple Ads, and timeout behavior. |
appstore-crash-insights | 69/494 | Triage TestFlight crashes, beta feedback, hangs, disk writes, launches, and performance diagnostics with asc crash/feedback/diagnostics commands. |
appstore-id-resolver | 65/318 |
pixijs
| Skill | Tokens | Description |
|---|
pixijs | 73/690 | Use first for PixiJS v8 tasks. Router for Application/app.init, scene graph Container/Sprite/Graphics/Text/Mesh, Assets, events, Ticker, filters, shaders, performance, migration, and create-pixi. |
pixijs-accessibility | 63/199 | Use for PixiJS v8 accessibility: AccessibilitySystem, screen reader overlays, keyboard navigation, accessibleTitle, accessibleHint, tabIndex, roles, activation settings. |
pixijs-application | 73/395 | Use for PixiJS v8 Application setup, app.init, renderer/canvas/screen/stage, resizeTo, ticker/sharedTicker, CullerPlugin, app.start/stop/destroy, releaseGlobalResources. |
pixijs-assets | 71/398 | Use for PixiJS v8 Assets: Assets.init/load/add/unload, bundles, manifests, cache, onProgress, background loading, spritesheets, video, SVG, fonts, compressed textures, parser selection. |
pixijs-blend-modes | 76/220 | Use for PixiJS v8 blend modes and compositing: normal/add/multiply/screen/erase/min/max, advanced-blend-modes, overlay, color-burn, hard-light, alpha behavior. |
pixijs-color | 70/256 | Use for PixiJS v8 Color: hex/CSS/rgb/hsl inputs, toHex/toNumber/toArray/toRgbaString, multiply, premultiply, alpha, tint, color-space conversion. |
pixijs-core-concepts | 68/243 | Use for PixiJS v8 renderer concepts: Application/Renderer relationship, WebGL/WebGPU/Canvas selection, render loop, systems, pipes, environment adapters, renderer fallback. |
pixijs-create | 69/249 | Use for PixiJS v8 project scaffolding with create-pixi: npm/yarn/pnpm/bun create commands, Vite templates, React template, non-interactive flags, existing project setup. |
pixijs-custom-rendering | 73/266 | Use for PixiJS v8 custom rendering: Shader.from, GlProgram/GpuProgram, UniformGroup typed uniforms, textures as resources, custom Filter, batchers, WebGL/WebGPU shader code. |
pixijs-environments | 70/230 | Use for PixiJS v8 environments outside normal browser pages: Web Workers, OffscreenCanvas, Node/SSR, CSP, DOMAdapter, BrowserAdapter, WebWorkerAdapter, unsafe-eval. |
tauri
| Skill | Tokens | Description |
|---|
tauri-config-security | 68/658 | Configure or review Tauri 2 tauri.conf files, capabilities, permissions, CSP, scoped filesystem/network/shell access, window labels, plugin permissions, and security-sensitive frontend API exposure. |
tauri-debug-testing | 77/577 | Debug, test, or stabilize Tauri 2 apps, including Rust compile/runtime errors, frontend API mocks, permissions failures, dev/build mismatches, WebDriver, CI checks, logs, DevTools, and platform-specific test gaps. |
tauri-distribution-mobile | 82/470 | Package, sign, notarize, update, release, or validate Tauri 2 desktop/mobile builds, including bundle targets, updater signatures, macOS/Windows/Linux distribution, Android/iOS setup, CI release checks, and store-readiness planning. |
tauri-ipc-plugins | 77/465 | Implement or review Tauri 2 Rust commands, frontend invoke wrappers, events, Channels, custom errors, state, official plugins, or custom Tauri plugins across Rust, JavaScript, permissions, and mobile surfaces. |
tauri-projects | 98/605 | Start, inspect, scaffold, migrate, or orient Tauri 2 projects. Use when a task mentions creating a Tauri app, adding Tauri to an existing frontend, src-tauri project structure, package-manager selection, framework selection, Tauri 1 to 2 migration, or "what shape is this Tauri repo in? |
tauri-shell-ui | 79/460 | Build or review Tauri 2 desktop shell features: windows, webviews, menus, tray icons, custom titlebars, resources, icons, state, sidecars, opener/shell APIs, deep links, and native-feeling desktop interactions. |
scientific-research
| Skill | Tokens | Description |
|---|
scientific-research | 91/2,024 | Use for scientific or scholarly research with source traceability, literature reviews, paper discovery, arXiv/OpenAlex/Crossref/Europe PMC/Semantic Scholar/PubMed queries, corpus building, DOI deduplication, source-backed claim extraction, evidence synthesis, or research quality validation. |
context-density
| Skill | Tokens | Description |
|---|
context-density | 137/2,806 | Use when designing, auditing, refactoring, measuring, or validating token-efficient context and prompt/output contracts. Trigger for AGENTS.md, prompts, skill packages, marketplace plugins, MCP/tool schemas, long-context placement, prompt/context compression, retrieval provenance, runtime context diagnostics, duplicated or overlapping prose, competing skill descriptions, companion-artifact drift, current-state commitments, irrelevant-context pruning, prompt reformatting, agent/subagent context handoffs, brittle trigger descriptions, strict JSON/schema/tool-call output, validators, retry/repair loops, or behavior-preserving token reduction. |
capability-workbench
| Skill | Tokens | Description |
|---|
agent-guidance-factory | 119/678 | Create, refresh, audit, or migrate repository agent guidance files such as AGENTS.md, AGENTS.override.md, CLAUDE.md, .claude/rules, and Cursor rules. Use when a repo needs durable coding-agent instructions, instruction load-order decisions, nested guidance, or cleanup of bloated or stale agent docs. Do not use for ordinary human README or CONTRIBUTING docs unless they must feed agent guidance. |
capability-auditor | 114/1,656 | Use when auditing agent skills or plugins for safety, evidence completeness across multiple artifacts, duplicated logic, token/context cost, prompt-contract risk, validation readiness, marketplace visibility, dependencies, network behavior, or install/update risk. Not for line, branch, mutation, or test coverage metrics. For structural split/merge/move/delete/router decisions across skills or plugins, use capability-portfolio-architect instead. |
capability-portfolio-architect | 103/947 | Use when auditing agent skill/plugin portfolio architecture for cross-plugin capability overlap, weak routing, duplicated or missing skills, split/merge/delete/move decisions, shared capability extraction, reference/script extraction, or plugin boundary redesign. For safety, coverage, or validation review of a single artifact without boundary changes, use capability-auditor instead. |
capability-reality-repair | 92/836 | Repair stale or false agent capability instructions immediately. Use when a skill, script, plugin manifest/helper, MCP/tool contract, connector guidance, validator, or bundled reference describes commands, schemas, paths, outputs, install state, dependencies, or behavior that does not match live reality. |
capability-synthesizer | 126/2,228 | Use when performing broad external-first discovery, vetting, scoring, distillation, or synthesis of agent skills or plugin packs from the public web, GitHub/public repositories, OpenClaw/ClawHub, marketplaces, research sources, community implementations, user references, or local skills. Use for well-vetted capability synthesis, cross-skill comparison, plugin-pack synthesis, global capability acquisition, skill strengthening, and adoption/rejection changelogs. |
capability-workbench | 74/1,768 | Route and orchestrate agent skill and plugin lifecycle work. Use when work involves agent skills, plugins, marketplace packages, capability acquisition, synthesis, creation, installation, packaging, audit, or context-density optimization. |
codex-cli
| Skill | Tokens | Description |
|---|
codex-cli | 91/1,048 | Route Codex CLI work across local CLI inspection, non-interactive exec/review automation, deferred completion without polling, live task and thread supervision with canonical checkpoint adoption guardrails, plugin and MCP management, doctor/debug/sandbox/app-server diagnostics, session log forensics, and Codex app local environment actions. |
codex-deferred-completion | 78/1,031 | Use when a long-running executable already publishes an atomic JSON terminal receipt and repeated native-session or remote-status polling would otherwise consume model turns; reserve a private receipt, launch the producer directly, and await it once. |
codex-doctor-debugger | 82/839 | Use when diagnosing Codex CLI install, config, auth, runtime, feature flags, sandbox denials, debug models, prompt input, app-server, remote-control, remote websocket/unix connections, or local Codex health failures. |
codex-environments | 84/995 | Create, update, debug, or remove Codex app local project environments and actions, including .codex/environments/environment.toml, Run/Test/Preview buttons, startup commands, simulator/browser/dev-server launch actions, environment variables, and repeatable local commands. |
codex-exec-automation | 97/1,018 | Use when preparing, running, debugging, or reviewing non-interactive Codex CLI automation with codex exec, codex exec resume, codex review, JSONL events, output schemas, last-message files, cwd/profile/config flags, sandbox and approval modes, or CI-style agent runs. |
codex-log-reader | 93/1,279 | Use when locating, summarizing, auditing, or debugging Codex session rollout JSONL logs by CODEX_THREAD_ID, cwd, query, issue key, project path, child or inherited-history symptoms, malformed or huge log symptoms, permission concerns, or "what happened in this Codex thread" questions. |
codex-plugin-mcp-manager | 77/895 | Use when managing Codex CLI plugins, plugin marketplaces, local marketplace refreshes, cache visibility, MCP server list/get/add/remove/login/logout, MCP bearer token environment variables, or plugin/MCP installation failures. |
codex-thread-supervisor |
scheduled-automation
| Skill | Tokens | Description |
|---|
scheduled-automation-runtime | 103/1,158 | Use when a local launchd, systemd timer, cron, or Windows Task Scheduler job works manually but fails or differs under the scheduler, or when registration, runtime context, last-result evidence, missed runs, or a fresh scheduler-originated result needs proof. Not for vendor CLI command construction, architecture inventory, cloud schedulers, or job business logic. |
gitlab-review
| Skill | Tokens | Description |
|---|
gitlab-review-response | 117/1,308 | Use when addressing existing GitLab merge request review discussions: classify feedback against current code and the latest diff, prepare focused fixes, prove exact source-head and CI provenance, and post idempotent same-thread replies. Supports plan-only, reply-only, and explicitly authorized per-thread resolution through GitLab REST v4 or glab api. Do not use for broad code review, GitHub pull requests, approvals, merges, or bulk resolution. |
stacked-delivery
| Skill | Tokens | Description |
|---|
stacked-change-delivery | 123/2,100 | Use when pull requests, merge requests, patch sets, or branch changes form a dependency stack and a child must remain based on an exact parent head. Covers stacked diffs, restacking after ancestor updates, per-node CI proof, safe bottom-up or forge-native atomic-prefix landing, isolated worktrees, and cross-agent handoffs. Do not use for one independent change, ordinary review-thread response, or automatic merge or force-push authorization. |
git-worktree-safety
| Skill | Tokens | Description |
|---|
git-commit-signing-recovery | 223/1,455 | Use when an ordinary Git commit fails before ref advancement at SSH signing with signer, agent, socket, helper-program, or signing-key evidence. Preserves the staged state, records private recovery state, issues one single-use signed retry authorization after a relevant signer-state change, and verifies the resulting commit, parent, diff, signer identity, and residual worktree state. Do not use for hook failures or hook-mutated state, conflicts, index or object corruption, remote authentication, OpenPGP or X.509 commit signing, editor/template or cleanup-transformed commit messages, artifact or release signing, key generation or rotation, initial or root commits, amend, merge, empty commits, index-altering -a/pathspec/interactive commit modes, non-signer command-scoped Git configuration or commit-metadata environment overrides, rebase/cherry-pick sequencers, unsigned-policy changes, or ordinary commit creation. |
git-worktree-recovery | 159/1,014 | Use when an expected Git worktree path or convenience symlink is missing, stale, or broken and a registered replacement may already hold the branch. Classifies branch-ref, reflog-only, object-only, or missing retention and guards POSIX-only exact symlink repair when directory-relative primitives are available. Do not use for Git administrative worktree repair after moving a main or linked worktree, ordinary worktree creation/removal/pruning, ref restoration, checkout/reset, stacked-change restacking, host-specific session orchestration, recovery of unsaved content from a vanished checkout, or arbitrary non-Git symlink repair. |
claude-code
| Skill | Tokens | Description |
|---|
claude-agent-worktrees | 90/756 | Use when coordinating Claude Code background agents, claude agents --json, dispatched-session defaults, git worktrees, tmux/iTerm panes, resume/continue/from-pr/fork-session, session names and IDs, remote control, prompt suggestions, or cloud ultrareview runs. |
claude-code | 72/789 | Route Claude Code CLI work across local CLI inspection, interactive and print-mode automation, plugin and MCP lifecycle, diagnostics, hooks/settings, background agents, worktrees, sessions, remote control, and ultrareview. |
claude-doctor-debugger | 91/738 | Use when diagnosing Claude Code install/update/auth/config/runtime health, broken customizations, safe mode, bare mode, debug logs, auto-mode classifier, settings validation, keychain or API-key auth boundaries, IDE/Chrome integration, doctor warnings, or update/install/setup-token issues. |
claude-hooks-settings | 87/708 | Use when creating, editing, auditing, or debugging Claude Code settings, hooks, CLAUDE.md/rules loading, custom agents, tool allow/deny permissions, output styles, workflows, setting sources, safe-mode/bare-mode differences, or plugin-provided customizations. |
claude-plugin-mcp-manager | 95/838 | Use when managing Claude Code plugins, plugin marketplaces, plugin validation/details/token cost/tag/install/update/remove/prune, session-only plugin-dir or plugin-url, MCP list/get/add/remove/import/serve, project MCP approvals, strict MCP config, transports, headers, OAuth, or MCP/plugin installation failures. |
claude-print-automation | 93/840 | Use when preparing, running, or debugging non-interactive Claude Code with claude --print, output-format text/json/stream-json, input-format text/stream-json, JSON schema validation, budget caps, fallback model, no session persistence, prompt suggestions, or trusted-directory CI-style runs. |
architecture-intelligence
| Skill | Tokens | Description |
|---|
architecture-conformance | 61/546 | Use when intended architecture must be compared with implementation: dependency policies, ADRs, ownership rules, drift, erosion, recovered models, and conformance classifications. |
architecture-decisions | 59/383 | Use when structural code choices need durable rationale: ADRs, tradeoffs, consequences, reversibility, ownership, review triggers, and validation plans. |
architecture-fitness-functions | 67/625 | Use when architecture choices need executable guardrails: dependency rules, cycle checks, boundary tests, ownership review gates, ADR compliance, runtime/resilience checks, and CI fitness functions. |
architecture-intelligence | 67/1,199 | Use whenever code work has structural consequences: project architecture, module boundaries, dependency direction, runtime topology, asynchronous state consistency, ownership, ADRs, fitness functions, or architecture refactoring. |
architecture-ownership-topology | 66/583 | Use when architecture crosses ownership or review boundaries: CODEOWNERS/OWNERS, module coverage, cross-owned dependencies, socio-technical coordination, and governance paths. |
architecture-refactoring-strategy | 67/517 | Use when code changes need staged structural design: boundary extraction, modularization, dependency inversion, migrations, anti-corruption layers, tests, rollout gates, and rollback. |
architecture-runtime-topology | 67/579 | Use when code work touches runtime shape: services, app/CLI/background flows, deployment/IaC, observability, resilience, external integrations, ownership, and runtime coupling. |
async-state-consistency | 118/2,314 | Use when designing, reviewing, debugging, or testing asynchronous state holders, caches, observers, subscriber notifications, memoized or coalesced loads, replay, or one-shot reads where invalidation, late subscribers, stale in-flight completion, read or publish races, out-of-order completion, or keyed/global ownership can expose stale state. Not for UI-only loading presentation, deployment topology, distributed consensus, or unrelated test flakiness. |
codebase-architecture-audit |
design-intelligence
| Skill | Tokens | Description |
|---|
design-intelligence | 69/852 | Route UX/product judgment to focused skills for framing, IA, interaction, usability, accessibility, visual communication, design-system governance, and heuristics. Not for Figma, CSS, automation, or assets. |
design-system-governance | 72/764 | Govern reusable design decisions, patterns, components, accessibility proof, contribution models, maturity, drift, and system adoption. Do not use for CSS implementation, Figma libraries, or token tooling unless requested. |
interaction-design | 64/768 | Design or critique flows, task sequences, affordances, feedback, states, error prevention, recovery, undo, progressive disclosure, input burden, keyboard/touch behavior, and interaction psychology. |
interface-architecture | 58/748 | Design or critique information architecture, navigation, taxonomy, labeling, screen structure, content priority, findability, search and browse strategy, and information hierarchy. |
product-framing | 68/505 | Frame product and UX decisions before interface work: strategy, user needs, Jobs-to-be-Done, outcome mapping, opportunity-solution trees, HEART/GSM, problem statements, assumptions, discovery plans. |
usability-accessibility-review | 71/742 | Review screens, flows, specs, or product ideas for usability heuristics, cognitive walkthrough failures, WCAG/APG/COGA accessibility risks, inclusive design, ethical UX, and dark patterns. |
visual-communication | 100/1,020 | Use when reviewing UI screenshots, golden images, visual diffs, or visual hierarchy for scannability, emphasis, readability, grouping, density, contrast, and perceptual clarity, especially when capture state or test-harness artifacts may be mistaken for product defects. Do not use for screenshot generation/export, CSS recipes, Figma operation, or decorative styling. |
game-design-intelligence
| Skill | Tokens | Description |
|---|
game-design-intelligence | 91/617 | Route source-backed game design work. Use for gameplay systems, core loops, progression, rewards, economies, balance, motivation, retention, onboarding, difficulty, multiplayer, live-service, content cadence, and player psychology. Do not use for engines, graphics, assets, programming, or implementation details. |
gameplay-systems | 97/379 | Design and critique core gameplay loops, mechanics, verbs, systems, dynamics, emergence, and player-facing structure. Use when the task is about what players do, why it stays interesting, how systems interact, or how the loop creates the intended experience. Do not use for engine, rendering, graphics, assets, or code. |
motivation-retention | 86/541 | Analyze player motivation, psychology, engagement, retention, ethical commercial design, and long-term value. Use for motivation models, retention risks, player segments, habit loops, emotional engagement, commercial viability, and dark-pattern review. Do not use for manipulative retention optimization. |
multiplayer-live-service | 92/476 | Design and critique multiplayer dynamics, cooperation, competition, social systems, fairness, toxicity prevention, guilds, matchmaking, content cadence, seasons, events, battle passes, late-game loops, and live-service health. Do not use for networking implementation, backend architecture, graphics, assets, or code. |
onboarding-difficulty | 71/384 | Design and critique onboarding, tutorials, FTUE, teaching, skill ramps, difficulty curves, assist modes, accessibility of challenge, failure, and mastery progression. Do not use for UI implementation or code. |
progression-economy-balance | 99/519 | Design and critique progression, rewards, power curves, economies, currencies, sinks, sources, balance, pacing, and tuning hypotheses. Use for reward loops, upgrades, unlocks, economy pacing, dominant strategies, or balance reviews. Do not use for implementation, analytics instrumentation, or monetization dark patterns. |
kotlin-multiplatform
| Skill | Tokens | Description |
|---|
kmp-architecture | 68/932 | Design or refactor Kotlin Multiplatform module boundaries, source-set hierarchy, shared logic/shared UI split, platform APIs, interop boundaries, and cross-platform library choices. |
kmp-compose-ui | 73/1,242 | Build and repair Compose Multiplatform UI, state, navigation, external URI effects, resources, platform entry points, previews, performance, accessibility, and UI tests in Kotlin Multiplatform projects. |
kmp-data-layer | 96/2,178 | Design and review KMP data layers, repositories, source-of-truth, legacy storage migration, durable migration decisions, resurrection prevention, DTO/domain mapping, sync, offline-first behavior, persistence choices, error handling, shared/coalesced work, cancellation and admission races, causal receipts, threading, and API exposure. |
kmp-ecosystem-selection | 85/400 | Select Kotlin Multiplatform ecosystem libraries, services, and tools for persistence, networking, DI, navigation, logging, observability, testing, code quality, resources, images, docs, payments, and reference templates without forcing a stack. |
kmp-gradle-doctor | 81/1,904 | Diagnose and fix Kotlin Multiplatform Gradle, source-set, dependency, private dependency resolution or consumption failures, Android target, Compose plugin, KGP/AGP, testing, static-analysis, and CI issues. |
kmp-interop-bridges | 89/581 | Design and review KMP platform bridges, source-set placement, expect/actual, entry-point wiring, cinterop, Swift API readiness, SKIE, KMP-NativeCoroutines, KDoctor, XCFrameworks, and SwiftPM export. |
kmp-migration-release | 86/727 | Plan and execute Kotlin Multiplatform migrations and release gates, including AGP 9 Android-KMP, monolithic composeApp splits, CocoaPods to SwiftPM, cinterop, iOS framework integration, CI, publishing, and app-store readiness. |
kmp-performance-observability | 80/461 | Diagnose Kotlin Multiplatform performance and observability across Gradle build time, Kotlin/Native memory and GC, Compose Multiplatform jank, binary size, startup, runtime logging, and release-mode verification. |
kmp-production-governance |
spec-driven-development
| Skill | Tokens | Description |
|---|
sdd | 59/760 | Route Spec-Driven Development work across lightweight, Spec Kit, Kiro-style, OpenSpec-style, brownfield, bugfix, planning, implementation, and audit lanes. |
sdd-audit | 51/411 | Audit SDD artifacts, requirement-task traceability, surface selection, and completion evidence before implementation or final delivery. |
sdd-implement | 54/511 | Execute SDD task lists safely, update task status, handle spec drift, and verify completion with fresh evidence before claiming done. |
sdd-plan-tasks | 55/513 | Convert approved SDD specs into technical plans, design artifacts, contracts, quickstart validation, and traceable task lists. |
sdd-spec-kit | 58/574 | Use GitHub Spec Kit projects and commands, including constitution, specify, clarify, plan, tasks, analyze, implement, extensions, and presets. |
sdd-specify | 56/498 | Create or refine SDD requirements and specifications with explicit assumptions, non-goals, acceptance criteria, success metrics, and retrofit truth markers. |
engineering-hygiene
| Skill | Tokens | Description |
|---|
code-maintenance-audit | 120/908 | Code maintenance sweep for changed and adjacent code. Use when implementing, modifying, reviewing, or finishing code changes, to catch dead code, stale leftovers, unused imports/variables/types/functions/components, duplicate branches/components/classes, obsolete tests/fixtures, and safe local consolidation opportunities before claiming done. Applies across languages and frameworks. Excludes speculative architecture rewrites, style-only churn, and performance/algorithm optimization unless the user explicitly asks. |
provisioning-missing-tools | 100/678 | Use when work is blocked, slowed, or downgraded because a command, SDK, runtime, package manager, system package, driver, CLI, simulator, emulator, or test/profiling utility is missing, outdated, too weak, or not configured, especially when the user asked for end-to-end execution. |
ui-visual-audit | 153/835 | Skeptical visual QA workflow for rendered UI and screenshots. Use when building, modifying, reviewing, or verifying web/mobile/desktop UI; captures browser, simulator, emulator, or desktop screenshots; compares implementation to a design; or claims a UI feature is visually correct. Forces a second pass for unrelated visual anomalies beyond acceptance criteria: occlusion, clipping, overlap, blank boxes, broken icons, poor contrast, broken spacing/alignment, inconsistent padding or margins, wrong platform controls, mock-looking or domain-implausible data, responsive/i18n text overflow, and accessibility-visible defects. |
untangle-business-logic | 164/818 | Targeted refactoring workflow for tangled business logic in code. Use when working on code where business rules, UI/presentation, API/IO, persistence, platform quirks, concurrency, lifecycle, state transitions, or error policy are mixed together; when the same domain rule is implemented in several places; when a bug depends on event ordering or stale state; or when a feature works but the business meaning is spread across components/services. Complements code-maintenance-audit: this skill is for semantic logic boundaries, hidden invariants, duplicated business meaning, and behavior-preserving logic untangling, not generic dead-code cleanup or performance tuning. |
signature-map
| Skill | Tokens | Description |
|---|
signature-map | 87/810 | Generates and queries signatures.json, a lightweight declaration index for fast, low-token code navigation. Use when locating type/function/method declarations in a medium or large repository, before broad rg/grep scans, when opening code around a known symbol, or when a repo's signatures.json index looks stale or broken. |
Repository Design
Source, install state, and runtime cache stay separated:
plugins/
<plugin-name>/
.codex-plugin/plugin.json
.claude-plugin/plugin.json
skills/
references/
scripts/
assets/
.claude-plugin/
marketplace.json
scripts/
install-codex-plugins.py
token-report.py
validate-repository.py
docs/
ARCHITECTURE.md
QUALITY.md
The public source surface is:
plugins/<plugin-name>/ for editable plugin source;.codex-plugin/plugin.json for Codex metadata;.claude-plugin/plugin.json for Claude Code metadata;skills/ for focused agent entrypoints;references/ for longer ledgers, contracts, scorecards, and source notes;scripts/ for deterministic validators and helpers;assets/ for plugin media and icons.
Generated or machine-specific state stays out of commits:
.agents/ is generated local Codex marketplace state;~/.codex/plugins/cache/... is Codex runtime cache;~/.claude/... is Claude Code runtime state;research/, skill-synthesis/, docs/superpowers/, and tmp/ are local
working areas unless content is intentionally promoted into source.
Development Workflow
-
Edit plugin source under plugins/<plugin-name>/.
-
Keep Codex and Claude Code manifests aligned when plugin metadata changes.
-
Keep large evidence, ledgers, and operating contracts in references/, not
in hot SKILL.md files.
-
Run the repository validator:
python3 scripts/validate-repository.py
-
Preview install-affecting changes:
python3 scripts/install-codex-plugins.py --dry-run
-
If the change should be usable in local Codex, refresh and verify the target
plugin:
python3 scripts/install-codex-plugins.py --plugin <plugin-name>
python3 scripts/install-codex-plugins.py --plugin <plugin-name> --check-only
-
Commit source files only. Do not commit generated marketplaces, caches,
bytecode, dependency folders, local research corpora, credentials, or
machine-specific paths.
Publication Standard
Before publishing or releasing a change, the repository should pass:
python3 scripts/validate-repository.py
python3 scripts/install-codex-plugins.py --dry-run
Publication-ready changes should be readable without local context, installable
from a fresh clone, and free of private project names, credentials, local
absolute paths, and runtime cache output. Docs should describe the actual
install model, and every listed plugin should have valid Codex and Claude Code
metadata.
See docs/ARCHITECTURE.md for the repository model,
docs/QUALITY.md for quality gates,
CONTRIBUTING.md for contribution standards, and
SECURITY.md for sensitive-data handling.
License
MIT. See LICENSE.
