Agent Plugins by Carinya Digital

Full-service digital agency plugins for small and medium businesses — brand, content, product, design, search, and engineering, delivered through Claude Cowork, Claude Code, Cursor, or the agent you already use. Each plugin bootstraps its own instance profile and ships MCP connectors for the tools you already use.

Install in one command

In Claude Code, Claude Cowork, or Cursor (Settings → Plugins). Other agent providers can load individual skill files (see skills.sh below).

/plugin marketplace add <path-to-this-repo>

Restart, then run /<practice>:setup. It bootstraps a git-versioned instance workspace (config/, brand/, squads/) and recommends your first practice plugin. Full walkthrough: brand-creative/README.md.

Important

Every output is a draft for your review — not production-ready deliverables or code without review, not a substitute for qualified professional judgment. Agents and skills draft work products; you verify accuracy, brand fit, accessibility, security, and compliance before anything ships.

Who this is for

Primary audience: small and medium business owners who need the services of a full-service digital agency — brand, content, product, design, and search — delivered agentically rather than by a retainer team.

Run the work in Claude Cowork, Claude Code, Cursor, or your preferred agent provider. You do not need an engineering team to start. Install the practices that match the work in front of you; add architecture and engineering when you are shipping a product or website.

This catalogue is how agency service lines show up in your session: structured skills, living profiles, and drafts you review before anything goes live.

Plugins at a glance

Install the practice plugins that match your work. The first practice setup writes config/instance.json if absent.

Plugin	Best for	First command
brand-creative	Brand voice and visual identity	/brand-creative:setup
content-marketing	Editorial calendar, social curation, CMS seed drafts	/content-marketing:setup
design	Wireframes, live UX review, and UX design fix	/design:setup
search-optimisation	Keyword research, technical SEO audits, on-page review	/search-optimisation:setup
product-management	Product strategy, roadmap, specs, research, metrics, backlog, sprint cadence, validation	/product-management:setup
architecture	Solution design and ADRs	/architecture:setup
engineering	Solution Design (design), implementation, code review, QA, platform ops	/engineering:setup
ralph-loop	Self-referential delivery loops (ad-hoc, custom, engineering-delivery)	/ralph-loop-setup
skills-index	Install-aware skill router	/skills-index:find
plugin-management	Create/customize plugins + skill quality gates	/plugin-management:create-plugin
document-management	Scaffold, reorganise, score, and keep current the docs/ tree	/document-management:docs-setup

Which plugin first? After setup, the interview recommends a starting practice. Common paths:

If you are…	Install next
Standing up a new business or brand	brand-creative
Running a website, content, and social	brand-creative → content-marketing → design
Need to rank in search	search-optimisation (+ content-marketing for on-page work)
Shaping a product or offering	product-management (+ design for new UI)
Shipping a website or app codebase	product-management + architecture + engineering (+ design for new UI)

Worked examples

Each example produces a draft artefact for your review — run the command, then verify assumptions, numbers, and brand fit before anything goes live.

1. Bootstrap a business instance

You have: a new business — name, one website, no instance repo yet.

Run: /brand-creative:setup --quick (or any practice setup) — answers business name and writes config/instance.json if absent, then the practice interview.

You get: a bound instance profile and a practice profile ready for the next skill.

2. Write a brand voice (Brand)

You have: a bound instance and some existing copy — website, social, or a voice interview.

Run: /brand-creative:brand-voice write — the skill reads seed material and drafts brand/brand-voice.md.

You get: a voice guide other content and product skills consume, ready for your review before anything publishes.

3. Draft this week's content calendar (Content)

You have: a brand voice on file and channels named in the content practice profile.

Run: /content-marketing:content-calendar write — name the window and any campaigns in flight.

You get: an editorial calendar with slot briefs — a draft for your review, not a scheduled publish.

4. Implement a UI task against design and AC (Engineering, when you ship code)

You have: approved {work-dir}/design.md, TASKS.local.md with Gherkin AC, and a bound target repo.

Run: /engineering:implement CHK01-01 — the skill reads the target repo's own AGENTS.md / CLAUDE.md before changing code.

You get: implemented code on a feature branch, ready for /engineering:code-review and your normal PR workflow.

More commands and personas: Entry points by service · Extended persona catalog · Skill & command reference.

---

Entry points by service

Personas are agency hats that map to slash commands inside practice plugins — not separate agent plugins. An owner can wear any of them. Install the plugin, run its setup, then invoke the command.

Brand, content & growth

Role	What it does	Command
Brand Lead	Voice lifecycle and visual identity guide	/brand-creative:brand-voice write
Content Strategist	Editorial calendar and social inventory curation	/content-marketing:content-calendar write
Content Writer	Blog posts, recipes, captions for CMS import	/content-marketing:draft-post
SEO Specialist	Keyword research, technical audits, on-page review	/search-optimisation:keyword-research
UX Designer	Wireframes, live UX review, UX fixes	/design:wireframe

Product & delivery

Role	What it does	Command
Product Manager	Product strategy, roadmap, specs from problem statements	/product-management:product
Delivery Lead	Tasks/backlog decomposition, sprint planning, validation	/product-management:sprint-planning
Architect	System architecture and ADRs	/architecture:solution
Engineer	Implement tasks against approved design and AC	/engineering:implement
Reviewer	Peer code review against design docs and AC	/engineering:code-review
QA	QA deploy, automated suite, exploratory pass	/engineering:exploratory-pass

Engineering skills share one library in engineering. Architecture (solution, adr) lives in architecture. Seniority labels reflect review depth, not separate plugins.

Run each plugin's setup before first use — every skill reads your instance profile and practice profile. Skipping setup is the most common reason output stays generic.

Extended persona catalog

Each persona below is named for the job it does. Start with entry points by service, then tune the underlying skill, practice profile, and connectors to how your business works.

Persona	What it does	Plugin	Command
Voice Enforcer	On-brand copy check against brand voice	brand-creative	/brand-creative:brand-voice enforce
Visual Identity Author	Colors, type, logo, UI tokens	brand-creative	/brand-creative:brand-guide write
Media Analyst	Vision analysis — subjects, season, mood, quality	content-marketing	/content-marketing:analyse-media
Caption Writer	Caption variants and channel copy	content-marketing	/content-marketing:write-captions
Technical SEO Auditor	Production audit → tracked issues	search-optimisation	/search-optimisation:technical-seo-audit
Backlog Owner	Epic breakdown, Now-phase scope, delivery risks	product-management	/product-management:tasks --product
Task Decomposer	Gherkin acceptance criteria per epic	product-management	/product-management:tasks
Epic Validator	Final sign-off against AC and roadmap gates	product-management	/product-management:validate
Research Synthesizer	Themes from interviews, surveys, and tickets	product-management	/product-management:synthesize-research
Competitive Analyst	Competitive analysis brief	product-management	/product-management:competitive-brief
Spec Writer	Feature spec or PRD from a problem statement	product-management	/product-management:write-spec
ADR Author	Architecture decision register and ADR files	architecture	/architecture:adr write
Epic Designer	Work-item technical design (TDD)	engineering	/engineering:design
MR Author	Merge request description from the branch	engineering	/engineering:merge-request
Docs Steward	Document-set quality and consistency review	engineering	/engineering:docs-review
Docs Author	Scaffold or reorganise docs/ to Diátaxis; score, drift, voice, and approved fixes	document-management	/document-management:docs-setup or /document-management:docs-improve
Debugger	Reproduce, isolate, diagnose, fix	engineering	/engineering:debug
Tech Debt Prioritizer	Prioritize remediation work	engineering	/engineering:tech-debt
WebOps Engineer	CI/CD, deployment, platform health	engineering	/engineering:platform-health

Everything here ships as Claude Cowork, Claude Code, or Cursor plugins, and individual skills can be installed via other agent providers that support skill files.

What's in the repo:

• Practice plugins — brand, content, product, design, search, architecture, and engineering — each with a setup interview, a living CLAUDE.md practice profile every skill reads, and propose profile update so conventions can be recorded as you work without re-running setup.
• Instance bootstrap — whichever plugin you install first writes config/instance.json if absent; no install-order dependency.
• MCP connectors — each practice ships an empty .mcp.json; add servers for your stack (source control, hosting, chat, trackers, analytics).
• Entry points by service — primary commands plus the extended catalog below.

Repository layout

brand-creative/           # brand voice + visual identity
product-management/       # product, roadmap, specs, research, metrics, backlog, sprint, validate
architecture/             # solution, adr
engineering/      # design, implement, review, QA, platform
design/           # wireframes, ux-design-review, ux-design-fix
content-marketing/        # calendar, curation, media analysis, CMS seeds
search-optimisation/      # keyword research, technical audit, content SEO review
ralph-loop/               # ralph-loop + ralph-loop-setup (+ hooks)
skills-index/             # install-aware skill router
plugin-management/        # create/customize plugins + skills-qa / skill-review
document-management/      # docs-setup, docs-improve (docs/ tree lifecycle)
scripts/                  # validate.py · validate_plugins.py · validate_skills.py
references/               # catalogue-only instance-profile-template + practice-setup-framework
.claude-plugin/marketplace.json   # plugin registry (name: agent-plugins)
.cursor-plugin/marketplace.json

Each practice plugin has the same shape:

<practice>/
  .claude-plugin/plugin.json
  .cursor-plugin/plugin.json
  CLAUDE.md               # template practice profile — filled in by /<practice>:setup
  README.md
  CONNECTORS.md           # category placeholders + suggested MCP providers
  .mcp.json
  skills/                 # skills — each is a /<practice>:<skill> slash command
  references/             # practice conventions
  hooks/                  # pre- and post-tool hooks (stubs today)

Getting started

First run (all surfaces)

1. Install a practice plugin from the marketplace (e.g. brand-creative or content-marketing).
2. Run /<practice>:setup — creates or binds your instance repo.
3. Install the practice plugins recommended during setup.
4. Run each practice's /<practice>:setup (e.g. /brand-creative:setup).
5. Bind targets — website pointer (config/target.json in target repos), credentials when ready.

Run practice setup first. Every other skill in a plugin reads from the profile it writes. The interview takes 10–20 minutes per plugin; --quick is available when you want to be productive in two minutes and refine later.

Claude Cowork

1. Open the Cowork tab.
2. Click Customize in the left sidebar.
3. Click Browse plugins and install from https://github.com/carinyadigital/agent-plugins, or upload a custom plugin (zip any practice directory).

After install, skills fire automatically when relevant; slash commands are available via /.

Claude Code

/plugin marketplace add carinyadigital/agent-plugins

/plugin install brand-creative@agent-plugins
/plugin install product-management@agent-plugins
/plugin install engineering@agent-plugins
/plugin install design@agent-plugins
/plugin install content-marketing@agent-plugins
/plugin install search-optimisation@agent-plugins
/plugin install ralph-loop@agent-plugins
/plugin install skills-index@agent-plugins
/plugin install plugin-management@agent-plugins

/brand-creative:setup
/product-management:setup

Updates: /plugin update.

Cursor

In Settings → Plugins → Add plugin:

• Paste this repo URL — https://github.com/carinyadigital/agent-plugins — then pick practice plugins from the marketplace list, or
• Upload a zip — zip any practice directory (e.g. engineering/) and drop it in.

skills.sh (skill files only)

Install individual skills without the full plugin surface (no hooks, MCP, or practice profiles):

# All skills from the monorepo
npx skills add carinyadigital/agent-plugins

# One skill
npx skills add carinyadigital/agent-plugins/engineering/skills/code-review

How it fits together

	What it is	Where it lives
Practice plugins	Self-contained service bundles — skills, hooks, MCP, and a template practice profile. Install the ones you need.	<practice>/
Skills	Domain expertise the agent draws on automatically — and slash actions you trigger explicitly: /brand-creative:brand-voice write, /content-marketing:draft-post.	<practice>/skills/<skill>/SKILL.md
Personas	Agency hats that map to skills — shared libraries inside each practice, not separate plugins.	Each practice's README.md
Instance profile	Git-versioned org facts, brand path, target bindings, squad charters.	<instance-repo>/config/instance.json, brand/, squads/
Practice profile	Per-practice conventions — stack defaults, persona preference, output formats, review gates.	~/.claude/plugins/config/digital-agency/<practice>/CLAUDE.md
Connectors	MCP servers that wire agents to your data — repos, hosting, design, trackers, chat.	<practice>/.mcp.json
Artefact consumption	Downstream practices read upstream outputs by path — brand voice, brand guide, wireframes — without hard install dependencies.	Resolved via instance/target pointers

Everything is markdown and JSON. No build step.

Delivery artefacts (target repo)

Strategy and delivery skills write into the bound target repo, not the plugin. Hierarchy is fixed:

Product → Solution → Roadmap → Backlog → work-item design / tasks → review → validate.

Stage	Skill	Writes
Product	/product-management:product	docs/product/product.md
Solution	/architecture:solution	ARCHITECTURE.md
Roadmap	/product-management:roadmap	docs/product/roadmap.md (does not require backlog)
Backlog	/product-management:tasks --product	docs/product/backlog.md (or tracker epics)

Default layout the skills expect:

docs/product/                 product.md, roadmap.md, backlog.md
ARCHITECTURE.md               system architecture (arc42)
docs/decisions/               register.md, ADR-NNNN-*.md
specs/{work-short-name}/          design.md; TASKS.local.md when required
specs/{work-short-name}/reviews/  code-review-{nn}.local.md + matching .local.json;
                                 ux-design-review-{nn}.local.md
docs/work/sprint-{id}/        plan.md, retrospective.md
reviews/                      ux-design-review.local.json,
                              review-learnings.local.md (gitignored — never committed).
                              code-review does not write here.
docs/reviews/                 agent byproducts (competitor-scan, metrics, digests)
TASKS.local.md                repo-root tracker-pointer cache (gitignored locally)
.ux-review/                   UX capture scratch (agent-local)
.claude/loop/                 Ralph run state on Claude Code (or .cursor/loop/ on Cursor)

Canonical detail: product-management/references/delivery-conventions.md.

Practice plugins by service line

Grouped by agency service line. Each plugin's setup is what tailors it to your business — start there.

Brand, creative & content

Plugin	What it adds
brand-creative	Brand voice lifecycle (discover, write, enforce) and visual identity guide (colors, type, logo, UI tokens). Writes to instance brand/ when bound.
content-marketing	Editorial calendar, social curation, media analysis, captions, and CMS seed drafting (posts and recipes). Two personas (Content Strategist, Content Writer). Reads brand voice from resolved brand path.
design	Wireframes, live-browser UX design review, and UX design fix. Writes to instance design/ when bound; other practices read wireframes via artefact consumption.

Product & strategy

Plugin	What it adds
product-management	Product strategy, roadmap, feature specs and PRDs, research, competitive briefs, metrics, stakeholder updates, backlog/tasks decomposition, sprint planning and retro, and work-item validation. Personas: Product Manager and Delivery Lead.

Growth & search

Plugin	What it adds
search-optimisation	Keyword research, production technical SEO audits, and on-page content SEO review. One persona (SEO Specialist). Optional pairing with content-marketing for seed review.

Engineering & platform

Plugin	What it adds
architecture	Solution design (solution) and architecture decision records (adr). One persona (Principal Architect). Companion to engineering for work-item design and implementation.
engineering	Solution Design (design), implementation, code review, merge requests, documentation passes, debugging, tech debt, QA deploy and exploratory validation, platform health. Five personas share one library; discover / deliver agents cover the usual path; architecture and product-management are recommended companions.

Platform

Plugin	What it adds
ralph-loop	Self-referential delivery loops — ad-hoc, custom, and engineering-delivery presets. Ships hooks.
skills-index	Install-aware skill router — /skills-index:find
plugin-management	Create/customize plugins, component authoring, marketplace registration, and skill quality gates (skills-qa, skill-review)

Companion practices: content-marketing invokes /product-management:tasks --product and /product-management:synthesize-research; search-optimisation invokes /product-management:competitive-brief rather than bundling duplicates. architecture invokes /engineering:design and /engineering:docs-review for work-item design and doc quality. engineering invokes /architecture:solution / /architecture:adr and /product-management:tasks for architecture and planning cadence. No direction requires the companion installed — skills degrade gracefully and document the pairing.

MCP connectors

Each practice plugin ships an empty .mcp.json. Add servers for your stack; skills produce usable output when no connector is configured.

Practice	Suggested categories
brand-creative	meeting transcription, knowledge base, design
product-management	project tracker, product analytics
content-marketing	creative / design, CMS, social scheduling
design	design, browser automation
search-optimisation	SEO intelligence, browser automation
architecture	source control (ADR harvest), project tracker
engineering	source control, browser automation, framework docs

Co-install companion practices or edit .mcp.json to add Slack, Notion, Vercel, Sentry, and other common servers — see each practice's CONNECTORS.md for the full placeholder map and suggested additions.

Connectors marked "customer subscription" need your own account and API key. Configure them in each plugin's .mcp.json or via claude mcp in Claude Code.

Instance bootstrap and practice profile

Two layers of configuration tailor generic skills to your business:

Layer	Path	Captures
Instance profile (shared, git-versioned)	<instance-repo>/config/instance.json, brand/, squads/	Business identity, target bindings, brand artefacts, squad charters
Practice profile (per plugin)	~/.claude/plugins/config/digital-agency/<practice>/CLAUDE.md	Stack defaults, persona preference, output formats, review gates, connector status

Run once per plugin you install:

Command	Writes
/<practice>:setup	Instance repo + handoff to first practice
/<practice>:setup	Practice profile for that service line

Framework: brand-creative/references/agency-setup-framework.md and catalogue references/practice-setup-framework.md (not copied into plugins).

Living profile. Every skill except setup uses propose profile update — show the exact diff, ask, write on yes. No skill auto-writes a full profile without confirmation.

Making it yours

These are reference templates. They get better when you tune them to how your business works — and the customization mechanism is the plugin itself.

• Run instance and practice setup. setup and setup are the customization mechanism. They interview you, read seed documents, and write profiles after you confirm the summary.
• Edit profiles directly. Instance facts live in your instance repo; practice conventions at ~/.claude/plugins/config/digital-agency/<practice>/CLAUDE.md. They survive plugin updates.
• Propose profile updates from any skill. When a stable convention surfaces as you work (tone corrections, channel mix, sprint length), skills show the exact change and ask before writing.
• Swap connectors. Point .mcp.json at your source control, hosting, design, and tracker stack. Skills fall back gracefully when a connector is not configured.
• Bring your brand and templates. Drop terminology, house style, and branded templates into the instance brand/ directory and practice profiles.
• Fork skills for house style. Every skill is a markdown file under skills/. Edit steps, gates, and output formats.

No build step. Everything is markdown and JSON.

Skill & command reference

The full map across all practice plugins. Run setup in each plugin before other commands.

brand-creative

Command	Skill	What it does
/brand-creative:setup	setup	Learns voice strictness, channels, seed material; writes practice profile
/brand-creative:brand-voice	brand-voice	discover, write, review, refine, enforce — brand/brand-voice.md
/brand-creative:brand-guide	brand-guide	write, review, refine — visual identity and UI tokens

product-management

Command	Skill	What it does
/product-management:setup	setup	Learns cadence, audiences, discovery, escalation, sprint length; writes practice profile
/product-management:product	product	write — docs/product/product.md (review via /engineering:docs-review)
/product-management:roadmap	roadmap	write — docs/product/roadmap.md after product (+ solution when present); does not require backlog
/product-management:write-spec	write-spec	Feature spec or PRD from a problem statement
/product-management:product-brainstorming	product-brainstorming	Sparring partner for ideas (no deliverable)
/product-management:synthesize-research	synthesize-research	Themes and insights from user research
/product-management:competitive-brief	competitive-brief	Competitive analysis brief
/product-management:metrics-review	metrics-review	Product metrics review with actions
/product-management:stakeholder-update	stakeholder-update	Status update tailored to audience
/product-management:tasks	tasks	--product → backlog after roadmap; {work-id} → specs/{work-short-name}/TASKS.local.md
/product-management:backlog-refine	backlog-refine	Groom backlog or judge sprint readiness
/product-management:sprint-planning	sprint-planning	Sprint plan — docs/work/sprint-{id}/plan.md
/product-management:sprint-retro	sprint-retro	Sprint retrospective — docs/work/sprint-{id}/retrospective.md
/product-management:validate	validate	Work-item completion sign-off against AC and roadmap gates

content-marketing

Command	Skill	What it does
/content-marketing:setup	setup	Learns channels, persona preference, seed sources
/content-marketing:content-calendar	content-calendar	write, review — editorial calendar and slot briefs
/content-marketing:curate-content	curate-content	Rank social inventory for upcoming posts
/content-marketing:analyse-media	analyse-media	Vision analysis — subjects, season, mood, quality
/content-marketing:write-captions	write-captions	Caption variants and channel copy
/content-marketing:edit-content	edit-content	Select or lightly edit best caption variant
/content-marketing:draft-post	draft-post	Blog post seed JSON for CMS import
/content-marketing:draft-recipe	draft-recipe	Recipe seed JSON for CMS import

design

Command	Skill	What it does
/design:setup	setup	Learns in-scope pages/flows and reference sources
/design:wireframe	wireframe	Low-fidelity layout and interaction spec from a brief
/design:ux-design-review	ux-design-review	Read-only UX review of implemented UI
/design:ux-design-fix	ux-design-fix	Address UX review findings or direct UI fixes

search-optimisation

Command	Skill	What it does
/search-optimisation:setup	setup	Learns target site, keyword themes, audit cadence
/search-optimisation:keyword-research	keyword-research	Topic keyword docs with intent and content opportunities
/search-optimisation:technical-seo-audit	technical-seo-audit	Production audit → tracked issues
/search-optimisation:content-seo-review	content-seo-review	On-page SEO review of content seeds

architecture

Command	Skill	What it does
/architecture:setup	setup	Learns target binding, architecture scope, companions
/architecture:solution	solution	write — ARCHITECTURE.md at the repo root (current or target); review via /engineering:docs-review
/architecture:adr	adr	plan, write, review — docs/decisions/

engineering

Command	Skill	What it does
/engineering:setup	setup	Learns stack, personas, target binding, connectors
/engineering:design	design	write — {work-dir}/design.md; review via docs-review
/engineering:discovery-review	discovery-review	Ready for Development gate on design.md + tasks
/engineering:implement	implement	Implement a task against approved design.md and AC
/engineering:code-review	code-review	Read-only peer review against design.md and tasks
/engineering:code-review-fix	code-review-fix	Address code-review findings without behaviour change
/engineering:merge-request	merge-request	Open merge request for implemented work
/engineering:merge-request-watch	merge-request-watch	Drive an open MR/PR to merge-ready
/ralph-loop:ralph-loop-setup	ralph-loop-setup	Seed and configure an autonomous delivery loop
/ralph-loop:ralph-loop	ralph-loop	Run an autonomous work-item delivery loop
/engineering:docs-review	docs-review	Read-only document-set quality and consistency review
/engineering:debug	debug	Bug investigation and fix
/engineering:tech-debt	tech-debt	Technical debt audit and prioritization
/engineering:deploy-qa	deploy-qa	Prepare QA workspace
/engineering:run-automated-suite	run-automated-suite	Run automated tests in QA workspace
/engineering:exploratory-pass	exploratory-pass	AC-driven exploratory validation
/engineering:document-defects	document-defects	Record defects from QA pass
/engineering:platform-health	platform-health	CI/CD, deployment, and platform health check

document-management

Utility plugin — owns the docs/ tree lifecycle. Does not replace /engineering:docs-review (set-quality review of any document set).

Command	Skill	What it does
/document-management:docs-setup	docs-setup	Scaffold a missing/stub docs/ tree, or reorganise a messy one to Diátaxis
/document-management:docs-improve	docs-improve	Score the tree, detect drift vs code, check voice; apply approved fixes (report-only never writes)

Contributing

Everything here is markdown and JSON. Fork, edit, PR. See CONTRIBUTING.md for design principles and the validation checklist.

• New skill → add <practice>/skills/<skill-name>/SKILL.md with name and description frontmatter. Invokable as /<practice>:<skill-name>.
• New persona row → add the skill under the owning practice and a row in that practice's README Agents/Personas table mapping the job title to the slash command.
• Validate before opening a PR — python3 scripts/validate.py lints manifests, verifies cross-file references, and validates evals schema. See AGENTS.md for repo conventions.

License

Licensed under the Apache License 2.0.

Copyright 2026 Carinya Parc Pty Ltd.