Working agent workflows,
installed in one paste.
superagnt is a marketing plugin for Claude and ChatGPT that helps businesses find customers through outbound, SEO and ads. These are its skills: Claude skills and Claude Code skills that run on live data out of the box, with nothing else to connect.
What is a Claude skill?
A Claude skill is a folder with a SKILL.md file in it: a name, a description of when to use it, and the instructions. Claude keeps only the name and description in context until a task matches, then loads the full file, and the reference files it names only when a step needs them.
Skills follow Agent Skills, an open format Anthropic released, so the same folder works in Claude Code, the Claude app, Codex and other agents that read SKILL.md files. A skill teaches the steps; it doesn't bring data. superagnt's skills install with the superagnt MCP server, so every step that needs LinkedIn, X, TikTok, YouTube, Reddit, search or people data can call it the moment you paste the install line.
Where skills live, and which one wins
| Scope | Path | Who gets it |
|---|---|---|
| Enterprise | The managed settings folder's .claude/skills/<name>/SKILL.md | Everyone in the organization |
| Personal | ~/.claude/skills/<name>/SKILL.md | You, in every project |
| Project | .claude/skills/<name>/SKILL.md | Everyone who works in that repository |
| Plugin | <plugin>/skills/<name>/SKILL.md | Anyone who installs the plugin |
Claude Code's skill locations, from Anthropic's docs as of Nov 3, 2026.
When two skills share a name, enterprise wins over personal and personal over project. Plugin skills load next to them, named /plugin-name:skill-name, so they never collide (Claude Code docs). In the Claude app, the skills you upload live in your account.
How to install a Claude skill
- Claude Code:
/plugin marketplace add superagnt/plugins, then/plugin install <skill>@superagnt. Add the superagnt connector to Claude first; it carries the data tools. - The Claude app: Customize, then Skills, then +, Create skill and Upload a skill, with the skill's folder as a ZIP. It needs code execution turned on (Claude help center, Nov 3, 2026).
- Codex:
codex plugin marketplace add superagnt/plugins, thencodex plugin add <skill>@superagnt. - Any agent that reads SKILL.md files:
npx skills add superagnt/plugins --skill <skill>. - ChatGPT and every other client: give your AI
superagnt.com/install; it carries the setup steps for that client.
Each skill's page below has its own install line, every client's snippet and every file it installs.
Every client side by side, with our timed installs and every failure we hit: How to install Claude skills and plugins in any client.
How to write your own skill
- Start SKILL.md with a frontmatter block:
nameanddescription. Claude Code treats every field as optional and only the description as recommended, because it is what Claude reads to decide when to use the skill. - Write the description in the words people type, saying what the skill does and when to use it. Claude Code cuts the description and its when-to-use text at 1,536 characters in its skill listing (as of Nov 3, 2026).
- Keep SKILL.md under 500 lines and move depth into reference files it names, so long material costs nothing until a step needs it.
- Test it by asking for the job in plain words, the way you would on a busy day, and watch whether the skill loads.
Skills, MCP servers, plugins and CLAUDE.md
| What it is | What it gives your AI | When it loads | |
|---|---|---|---|
| Skill | A SKILL.md folder | The steps for one job | Its description every session, the rest when a task matches |
| MCP server | A connection to outside tools and data | Something to call: an app, a database, a data source | Its tool list when you connect |
| Plugin | A bundle: skills, MCP servers, commands and hooks | One install for all of them | Each part as it would on its own |
| CLAUDE.md or custom instructions | Standing notes Claude reads | Rules and context for every answer | In full, every session |
| superagnt | A plugin: 10 marketing skills plus the superagnt MCP server | The steps and the live data they call, on one key | Skills when a task matches; data tools when you connect |
ChatGPT and Codex use the same split: OpenAI's plugins bundle skills and MCP servers, and they work in ChatGPT's Chat and Work and in Codex (OpenAI, Nov 3, 2026).
Is a third-party skill safe?
A skill can tell your AI to run commands or reach outside services, so read it before you install it. Anthropic's advice, as of Nov 3, 2026, is to install skills only from trusted sources and to review any other before enabling it: its files, scripts, code dependencies and any instruction to connect to an outside network (Claude help center). In Claude Code, a project skill's allowed-tools applies even in a folder you never trusted (as of Nov 3, 2026), so check that field in any skill a repository carries.
Every superagnt skill is plain markdown with no scripts, published in the superagnt/plugins marketplace on GitHub, and each one's page shows every file before you install it. The table below is built from those files.
| Skill | Files | Size | Turns on | Never |
|---|---|---|---|---|
| LinkedIn Warm Outbound | 9 | 113 KB | data:linkedin, database | starts a campaign, or sends without your say-so |
| Content Ideas Engine | 4 | 6.5 KB | data:x, database | posts, replies or DMs |
| Competitor Tracker | 4 | 7.7 KB | data:x, database | posts, or makes your scoreboard public |
| Competitor Ad Research | 4 | 19 KB | data:facebook, data:web, data:seo, database | creates, edits or runs ads, or connects an ad account |
| Outbound Pipeline Engine | 4 | 7.3 KB | data:linkedin, database | sends an email or a LinkedIn message |
| Super SEO | 17 | 495 KB | data:seo, data:web, database | publishes, sends outreach or spends money |
| Video Multiplier | 4 | 8.8 KB | data:youtube, database | uploads, posts or edits your YouTube studio |
| Hook Lab | 4 | 7.8 KB | data:tiktok, database | posts, comments or DMs |
| Newsletter Autopilot | 4 | 7.7 KB | data:x, data:web, database | sends, schedules or publishes |
| Agency Outbound | 4 | 24 KB | data:linkedin, database | starts a campaign or moves rows between clients |
Read from each skill's plugin folder and its SKILL.md when this page was built.
What 10 skills cost in context
An installed skill keeps its description in every session. Our 10 skills' descriptions come to 5,593 characters, about 1,398 tokens at 4 characters a token (an estimate, not a tokenizer count). The longest is LinkedIn Warm Outbound's at 875 characters, the shortest Content Ideas Engine's at 382. Installing any of them also installs the base superagnt plugin, once, and its 8 skills add 3,791 characters: all 10 with the base come to 9,384, about 2,346 tokens. The rest of each skill loads only when a task calls for it.
The same prompt, with and without the Competitor Tracker
Same model (Opus 5.5), same four brands, same question on Nov 3, 2026: how do Semrush, Ahrefs, Moz and SE Ranking stack up on X right now? Each side was a fresh Claude Code run, one with the Competitor Tracker skill and superagnt, one with only web search, web fetch and a shell.
| Claude Code on its own | With the skill | |
|---|---|---|
| What it reached | A login wall on X search, and blocked or shut-down mirrors | The latest 20 posts per brand from X search |
| Posts per hour | Empty for all four | Semrush 7.13, Ahrefs 4.45, SE Ranking 0.58, Moz 0.32 |
| Work | 17 tool calls, 4 min 39 s | 8 superagnt calls, 5 credits, 9 min 36 s |
The skill brought the method; the superagnt connection brought the data. The skill's run, call by call.
Claude skills: common questions
01Are Claude Skills free?
Skills are files, and Anthropic's help center lists skills on the Free, Pro, Max, Ultra, Team and Enterprise plans, with code execution turned on (as of Nov 3, 2026). superagnt's skills are free to install and read; their live data calls run on superagnt credits. Plans are $49, $149 or $399 a month, and a new account starts with a no-card trial.
02Are skills secure?
A skill can tell your AI to run commands or call outside services, so read it before you install it. Anthropic says to install skills only from trusted sources. Every superagnt skill is plain markdown you can read on its page, and none of them sends a message, publishes a post or buys anything without your say-so.
03How many skills can I install at once?
Anthropic's docs set no number, as of Nov 3, 2026. The limit is the skill listing: Claude Code keeps every skill's name in context and fits the descriptions into a budget of 1% of the model's context window, cutting each at 1,536 characters. Past the budget it drops the descriptions of the skills you use least, so Claude can miss when to use them. What our 10 cost is above.
04Do skills only work with Claude?
No. Agent Skills is an open format, so Codex and other agents read the same folders. OpenAI's plugins carry skills and MCP servers into ChatGPT's Chat and Work and into Codex (as of Nov 3, 2026). superagnt's skills install in Claude Code, the Claude app and Codex, and the install page sets them up in ChatGPT.
05What is the difference between skills and custom instructions or MCP?
Custom instructions, like CLAUDE.md, load in every chat and shape every answer. A skill loads only when a task matches and teaches one job step by step. An MCP server connects your AI to outside tools and data. A skill says how to do the job; MCP gives it something to do it with. superagnt ships both in one install.
06Where can I find community-made skills?
Anthropic's own skills repository and plugin marketplace, skill directories such as skills.sh, and GitHub. Ours are the 10 below, in the superagnt/plugins marketplace on GitHub. Read any skill's files before you install it.
10 Claude skills for marketing, with the data built in
LinkedIn Warm Outbound
This week's warm leads: the people engaging with posts about your problem, scored and staged with a first message each.
turns on data:linkedin, database · never starts a campaign, or sends without your say-so · 9 files, 113 KB
example run · fictional data
details
Your market tells you who's shopping, in public, every day: they react to and comment on posts about the problem. This agent finds those posts, pulls everyone on them, and scores each person against your ICP before it pays for a single email lookup. The ones who fit land in HeyReach, Instantly, AgentMail or your Gmail with a first message written from what they engaged with. You press start.
LinkedIn posts, reactions + comments · People + company enrichment (first-party) · Email finder + verifier (first-party) · Workspace Postgres · HeyReach · Instantly · AgentMail · Gmail
Content Ideas Engine
Wake up to 5 post ideas, each cited to a real conversation in your niche.
turns on data:x, database · never posts, replies or DMs · 4 files, 6.5 KB
example run · fictional data
details
Content ideas are a volume game, and the volume part is annoying. This agent runs 8-12 standing X searches every night, files what your audience actually talks about into your workspace database, and writes you an ideas report with the receipts linked. You read it with coffee and pick.
X search + threads · Workspace Postgres
Competitor Tracker
A daily scoreboard of you vs your competitors on X, on a chart you can share.
turns on data:x, database · never posts, or makes your scoreboard public · 4 files, 7.7 KB
example run · fictional data
details
You can feel like you’re winning on X and be wrong. This agent runs the identical search for every brand you track each morning, logs posts-per-hour and engagement medians to your workspace database, and keeps a shareable chart current. Same method every day, so the trend is real, and the numbers are yours to publish.
X search · Workspace Postgres · Workspace dashboard
Competitor Ad Research
Your competitors' running Facebook and Instagram ads, grouped by offer, with each landing page's price and what the offer's keywords cost per click.
turns on data:facebook, data:web, data:seo, database · never creates, edits or runs ads, or connects an ad account · 4 files, 19 KB
details
Your competitors' ads show which offers they pay to push. This agent pulls the ads they run from the Meta Ad Library, groups them by offer, angle and format, reads each landing page for its offer and price, and prices the offer's keywords. One report, every row kept in your workspace database, and it never touches an ad account.
Meta Ad Library (Facebook + Instagram) · Landing page reads · Keyword volume + cost per click · Workspace Postgres
Outbound Pipeline Engine
10 scored prospects with verified emails and outreach drafts, every morning.
turns on data:linkedin, database · never sends an email or a LinkedIn message · 4 files, 7.3 KB
example run · fictional data
details
Prospecting isn’t hard, it’s the same six steps every day: find the company, find the person, get the email, verify it, score it, draft the note. This agent runs all six against your ICP and keeps a pipeline table current with a draft per prospect. Sending stays yours.
People + company search (first-party) · Email finder + verifier (first-party) · LinkedIn data · Workspace Postgres
Super SEO
Every move your site could make this week, ranked on one scale, each with the Search Console number behind it.
turns on data:seo, data:web, database · never publishes, sends outreach or spends money · 17 files, 495 KB
example run · fictional data
details
Most SEO advice is a list with no order, so the cheap win and the six-month bet sit on the same line. This agent reads your Search Console first: which pages are losing clicks, which queries sit just off page one, what is not indexed at all. Then it scores every move it could make, drafting a page, rewriting a title, pitching a link, fixing an index problem, on one scale, and works the board from the top. You approve; it executes.
Google Search Console (your property) · SERP + keyword + backlink data · AI answer visibility · Web scrape · Workspace Postgres
Video Multiplier
One upload in, a text package out: chapters, five titles, a description and up to six post drafts.
turns on data:youtube, database · never uploads, posts or edits your YouTube studio · 4 files, 8.8 KB
example run · fictional data
details
You already did the hard part: you recorded the video. This agent pulls the transcript, writes the chapters, five title options and the description, then drafts up to six posts from the same material: an X thread, a LinkedIn post, a newsletter section and two or three short-form scripts. It writes text and never cuts clips. Publishing stays yours.
YouTube transcripts + video data · Workspace Postgres
Hook Lab
Ready-to-film scripts every week, each built on a hook that already won in your niche.
turns on data:tiktok, database · never posts, comments or DMs · 4 files, 7.8 KB
example run · fictional data
details
The first two seconds decide whether a video runs, and your niche already ran the test. This agent studies the top-performing videos each week, ranks the hook patterns that keep winning, and writes ready-to-film scripts in your voice: hook, beats, CTA, proof linked. You film; it queues next week’s batch.
TikTok search + video data · Workspace Postgres
Newsletter Autopilot
A finished issue draft every send day, researched all week, every claim cited.
turns on data:x, data:web, database · never sends, schedules or publishes · 4 files, 7.7 KB
example run · fictional data
details
The newsletter dies because the research eats the week. This agent sweeps X and the web for your topics every night, keeps only what is genuinely new, and on send day assembles a complete issue draft in your voice: sections, links, subject line options. You edit it and hit send. Sending stays yours.
X search + threads · Web search + scrape · Workspace Postgres
Agency Outbound
A list from the people engaging with your competitors' LinkedIn posts, pushed into the sequencer you already run, with replies and meetings synced back into a dashboard you share by link.
turns on data:linkedin, database · never starts a campaign or moves rows between clients · 4 files, 24 KB
details
Run outbound for every company you work with, one workspace each. This agent pulls the people reacting to and commenting on competitors' LinkedIn posts and on the hashtags their buyers follow, enriches them, keeps the ones who fit and stages them in that workspace's own HeyReach, Instantly, Smartlead or lemlist campaign. Replies and booked meetings sync back into the workspace's database, and the agent turns them into a dashboard you share by link, or a report in chat where dashboards aren't on yet.
LinkedIn company posts, hashtags, reactions + comments · People + company enrichment (first-party) · Email finder + verifier (first-party) · Workspace Postgres + vendor sync · Workspace dashboard · HeyReach · Instantly · Smartlead · lemlist · EmailBison · Expandi · La Growth Machine · Apollo