Skip to the guide

Guide

The Authority Hacker Skill Library, Mapped

  • claude-code
  • skills
  • plugins

Eighteen plugins install into Claude Code and unpack into thirty eight separate skills. Most people use four of them and forget the rest exist. Here is what each one does and when it is the right tool to reach for.

CountWhat
18Plugins installed
38Skills total
5Router based suites

The pattern · Routers triage, lanes execute

Five of these plugins use the same architecture. It is worth learning once because it tells you which command to actually type.

The router

A small skill that reads your request, works out what you are really asking for, loads any saved state, and hands off. It rarely produces the output itself. Examples are seo-audit, image-generation, email-marketing, topical-map and delegate.

The lanes

Specialist skills that hold the real depth. A lane assumes the router already gathered context, so calling one cold can work but you lose the setup. Examples are technical-seo-audit, gpt-image, email-copywriter and topical-map-ahrefs.

Category one · Content production

Five skills that write, rewrite, or reshape content for a channel.

Writing and rewriting

5 skills, all standalone.

  • blog-post-writer · Standalone. SEO article writer. It detects your brand setup, picks a format such as a list post, how to, comparison or review, runs SERP research, then drafts section by section from a defined set of section types. Reach for it when you want a full article rather than an outline.
  • content-humanizer · Standalone. Strips AI tells from a draft you already have, without touching meaning, links or voice. The interesting design choice is that it adjudicates scanner flags instead of obeying them, then verifies the rewrite. Reach for it when a draft reads templated. It will not write anything new.
  • newsletter-maker · Standalone. Six stages from brand setup through news discovery, story selection, drafting, your review, and HTML generation. Reach for it when you ship on a cadence. It is built for recurring issues, not one off sends.
  • social-content · Standalone. X, Threads and LinkedIn. Extracts from source material or interviews you, matches your voice, decides whether the post is worth saving to reusable memory, drafts, edits, adds image prompts and can hand off Typefully ready drafts. Reach for it when one idea needs to land on three platforms.
  • linkedin-carousels · Standalone. Turns source content into a LinkedIn document carousel with outline, slide copy, teaching panels, cover and CTA slides, PDF export and a companion post. The largest single skill in the library, with hard art direction rules and approval gates you cannot skip. Reach for it when an article deserves a second life as a carousel.

Eleven skills across three plugins. This is the deepest part of the library by a wide margin.

seo-audit

1 router, 5 lanes, 1 setup helper.

  • seo-audit · Router. The command center. Intake, then a command plan, then six stages of evidence gathering, then synthesis into a prioritised report. Depth presets let you scale the work to how deep you actually want to go. Reach for it when you want the whole picture rather than one angle.
  • technical-seo-audit · Lane. URL first technical pass. Crawlability, indexability, robots rules, sitemaps, status codes, redirects, canonicals, structured data, internal links and JavaScript renderability.
  • content-audit · Lane. Page level quality. Search intent fit, titles, meta descriptions, headings, internal links, content decay, experience and trust signals, thin pages and refresh candidates.
  • gsc-audit · Lane. Opportunity mining from your own Search Console data. Click through rate wins, decay, cannibalisation, dead pages, striking distance keywords and non brand visibility. The highest signal lane if your site already has traffic.
  • market-visibility-audit · Lane. The competitive lane. It normalises evidence from either Ahrefs MCP or DataForSEO MCP into keyword gaps, ranking pages, content hubs and traffic opportunities.
  • ai-visibility-audit · Lane. Answer engine readiness. AI crawler access, entity clarity, citation worthy pages, robots rules for AI agents, plus bounded live probes against ChatGPT Search and Google AI Search.
  • setup-gsc · Setup. Wires up the Google Search Console connection so the GSC lane has first party data to work with. Run this once, before the audit.

on-page-optimizer

1 skill, 6 stages.

  • on-page-optimizer · Standalone. One page against one keyword. Site setup, memory load and route, SERP analysis, gap analysis, findings, optimisation, then verification and a log entry. It remembers past cycles, edits Astro sites directly and runs a build check after changes. There is also a review-all mode for a whole site. Reach for it when the page exists and is underperforming.

topical-map

1 router, 2 provider lanes.

  • topical-map · Router. Picks your data provider and hands off. Use this unless you already know which provider you want.
  • topical-map-ahrefs · Lane. Builds the map from competitor top pages, then expands laterally into long tail, then runs a semantic cleanup. Has depth profiles and respects MCP row limits.
  • topical-map-dfseo · Lane. The DataForSEO Labs version. Topic clusters, pillar pages and keyword research, with quality gates, size presets and its own sub agents doing the heavy lifting.

Category three · Visuals

Seven skills. The image plugin is the clearest example of the router pattern in the whole library.

image-generation

1 router, 3 engine adapters, 2 helpers.

  • image-generation · Router. Decides which engine gets the job, enforces a dry run before anything costs money, and owns the shared prompt library the adapters all draw from.
  • gpt-image · Adapter. OpenAI API direct. Chosen when you need exact sizes, quality control, masks, specific formats, edits, batching or identity preservation. Needs an API key.
  • codex-image · Adapter. The same model routed through the Codex CLI, so it bills your ChatGPT plan instead of an API key. This is the fallback when OpenAI API setup is missing, and it is the cheaper path if you already pay for Codex.
  • nano-banana · Adapter. Gemini. Chosen for Gemini aspect ratios, 2K and 4K output, or when the GPT routes are unavailable.
  • infographic · Workflow. Guided path from source content to an infographic, cheat sheet, poster, diagram or data visual. It hands off to whichever engine fits the brief.
  • brand-extract · Helper. Pulls colours, fonts, logos and assets off a website so generated visuals match the brand. Quietly the most useful skill in the plugin, because it feeds all the others.

video-thumbnail

1 skill.

  • video-thumbnail · Standalone. YouTube thumbnails on gpt-image-2. Angles, headline text, visual layouts, reference assets, generation, then a file size optimisation pass. It carries an explicit cost callout before it spends anything.

Category four · Email and ads

Six skills. The email suite is a genuine pipeline where each stage produces a file the next stage consumes.

email-marketing

1 router, 4 pipeline stages.

  • email-marketing · Router. The front door. Triages what you asked for and routes it, and owns state bootstrap, checkpoints and the handoff contracts between the stages below.
  • campaign-planner · Stage 1. Setup, intake, sequence strategy, then your approval. It writes a strategy file that the copywriter reads. Built around launches, flash sales, webinars and deadline sequences.
  • email-copywriter · Stage 2. Turns an approved strategy, or a complete one off brief, into actual email bodies.
  • email-image-uploader · Stage 3. Pushes local image assets to stable public URLs so the email HTML does not break in the inbox.
  • bento-broadcast · Stage 4. Creates draft broadcasts in Bento from the Markdown or approved HTML. Note the hard rule baked into it. It drafts, and it never sends or schedules.

meta-campaign

1 skill.

  • meta-campaign · Standalone. Facebook and Instagram campaigns built from a product or an offer. Positioning, campaign strategy, ad copy, creative angles, image prompts and assets. It ships with an explicit model calibration block, which is unusual and worth reading before you run it.

Category five · Research, delegation and plumbing

Ten skills that never touch a deliverable but change how everything else runs.

delegate

1 router, 2 supporting skills.

  • delegate · Router. Splits work across models. It loads a saved routing profile, decides between a hybrid run and a Claude only run, fans the work out, then runs an independent review pass over the result. Its own trigger rules force it to fire on any list task of ten or more items and on any request for a critique.
  • clarify · Lane. Interviews you until the plan is genuinely locked. Built for fuzzy tasks that would otherwise be executed on a guess.
  • codex · Lane. The second engine. Second opinions, structured review, plan critique and hands on coding, in chat, review or rescue mode. It is explicitly barred from drafting marketing prose unless you override that, which is the right call.

Research and build

3 skills.

  • deep-research · Standalone. Runs long form research through Gemini Deep Research and returns a cited report. It is asynchronous, so you submit and then check status, wait, or ask a follow up by ID. Three depth tiers.
  • new-site · Standalone. Interview, brand brief, visual direction, site blueprint, a fresh Astro scaffold, then several homepage concepts generated in parallel. Phase one runs in plan mode so you lock the direction before any code exists.
  • cf-scraper · Standalone. For pages that normal fetch tools choke on. JavaScript heavy, protected or bot walled. Outputs markdown, screenshots, PDFs or structured data extracted by a custom prompt. Batch and crawl modes included.

Utilities

3 skills.

  • import-updates · Standalone. The one nobody knows they need until they have edited a skill. It does a three way merge between the version you forked from, the new version that shipped, and your customised copy. That ancestor is the whole point. Without it a plain diff mislabels your own edits as incoming changes. It backs up before any write and shows a plain English menu instead of conflict markers.
  • ah-feedback · Standalone. Sends bug reports, feature requests and interface complaints back upstream from inside the chat. It asks for approval before anything leaves your machine.
  • private-marketplace-setup · Standalone. Build your own private plugin marketplace for a team. It packages skills as plugins, generates the manifests, validates them, simulates what an installing user sees, writes the install docs and hosts the whole thing on GitHub. The most reusable skill here if you run an agency.

Getting value out of it · What most people get wrong

Counts reflect the plugins installed on my machine as of August 2026. The marketplace ships one more plugin I have not installed, so treat 18 and 38 as my inventory rather than the full catalogue.

Back to every guide