agentleFS
XPOZpublicXPOZpublic/geo-seo-agent/.claude/skills/setup/SKILL.md

setup

Skill0 starsChanged 3 months agoMITView on GitHub

First-run onboarding. Interview the operator about their company and product, write config.json and strategy/company-profile.md, check credentials and optional integrations, and initialize program state. Run once before anything else; re-run to change company facts or integrations.

  • Reads credentials
  • Installs packages

192 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use add-atomic-chat-tool, add-dashboard and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • MicrosoftSkill

    add-atomic-chat-tool

    Add Atomic Chat MCP server so the container agent can call local models served by the Atomic Chat desktop app via its OpenAI-compatible API.

    Also in 8 projects that use this

  • MicrosoftSkill

    add-dashboard

    Add a monitoring dashboard to NanoClaw. Installs @nanoco/nanoclaw-dashboard and a pusher that sends periodic JSON snapshots.

    Also in 8 projects that use this

  • +10 more from microsoft

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 169 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Brand terms: every name the brand appears as in text (company name, product names, common misspellings worth counting).

    line 17
  2. Distinctive:

    Domains: every domain that counts as "ours" when cited (main site, docs subdomains count automatically, separate product domains do not …

    line 18

“Common” is measured: kept by at least 6 in 10 of the 169 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Setup

Everything downstream hangs off two files this skill produces: config.json (what the scripts read: brand terms, domains, engine models) and strategy/company-profile.md (what the skills read: product facts, personas, problems solved, positioning). Until they exist, no other skill can run meaningfully. The interview is a conversation, not a form; ask follow-ups when an answer is vague, because a vague company profile produces vague prompts and vague content.

1. Interview the operator

Collect, in whatever order the conversation flows:

  • Company and product: name, what the product does, the category it competes in, pricing model, differentiators. Concrete facts, stated the way the company wants them expressed publicly.
  • Problems solved: the 3-7 problems a buyer has when they need this product. These seed /target-prompts-discovery.
  • Personas: who buys and who uses it (e.g. developer, marketer, founder). Discovery and the panel's Value ratings lean on these.
  • Brand terms: every name the brand appears as in text (company name, product names, common misspellings worth counting). These become the mention-detection regex, so ambiguous terms that also appear as common words will produce false positives; flag that and prefer distinctive terms.
  • Domains: every domain that counts as "ours" when cited (main site, docs subdomains count automatically, separate product domains do not unless listed).
  • Known competitors: whoever they already know they compete with. Just names; /target-prompts-discovery and /content-gap-analysis research entities into memory/competitors/ as evidence accumulates.
  • Publishing workflow: where produced content ends up (CMS, static site repo, marketing team handoff) and what the human does to publish. Recorded in the company profile; /content-generation follows it when defined and otherwise delivers drafts. If the workflow uses hero images, collect a brand style direction (style, colors) and aspect ratio into config.json's hero_images block; /generate-hero-image reads it.
  • Channels: all six standard channels are active from day one; this question collects each channel's specifics for the content plan's channel table. Public documentation: where the docs live (platform or repo, and the domain if separate from the main site), or n/a if the product has no docs. GitHub: the public org/repos (or n/a if the product has no public code). Integrations: what exists or is planned (MCP server, ChatGPT app, Claude Code plugin, Cursor plugin, framework integrations, Zapier), or n/a. Review sites: which sites matter in the category (G2, Capterra, Product Hunt are the usual starters). Mark a channel n/a only when it genuinely cannot apply, not because it's not ready yet. Point out the quick wins while you're at it: a first /site-readiness run plus a pass on review-site profiles and GitHub metadata is hours of work and immediately valuable, worth doing before the first content batch.
  • Engines: which of Claude (Anthropic API), ChatGPT (OpenAI API), and Gemini to track. Default to all three; each needs its credentials in .env (Gemini takes a plain GEMINI_API_KEY, or Vertex AI service-account credentials). Model ids default to the ones in config.example.json; let the operator override.
  • Integrations: all optional; the loop runs on engine keys and Xpoz alone. Google Search Console (the Google surface per prompt) and Google Analytics 4 (AI-referral traffic): if the operator has them and wants them measured, they connect the corresponding MCP servers in their Claude Code config; record the choice in config.json under integrations so skills know what to expect. Also recommend where relevant: BigQuery via an MCP server, for querying the GA4 and Search Console bulk exports past the API sampling limits, and, when the product has docs, a docs platform that serves llms.txt and per-page markdown (e.g. Mintlify), which strengthens the docs channel's extractability.

2. Write the outputs

  • config.json from config.example.json with the collected values.
  • strategy/company-profile.md following that file's template sections.
  • Fill the "What the company is" section of strategy/geo-strategy.md.
  • In strategy/content-plan/content-plan.md: fill the channel table's notes column (site and content types, review sites, org/repos, integrations, n/a markers) and record the setup decisions in the plan's change log.

3. Verify the environment

  • .env: many agent harnesses block reading and writing .env files entirely, so hand this step to the operator by default: ask them to copy .env.example to .env and fill in the keys for every configured engine, and tell them which keys those are. Don't try to read the file to verify; the smoke test below is the verification.
  • .mcp.json exists with the Xpoz MCP server (create from .mcp.json.example if missing). Xpoz powers community-demand research in /target-prompts-discovery; the operator needs an account at https://xpoz.ai, and most MCP clients handle OAuth automatically on first tool call.
  • Python environment: python3 -m venv .venv && .venv/bin/pip install -r scripts/requirements.txt.
  • Smoke-test one configured engine with a trivial question through its ask script (e.g. .venv/bin/python scripts/ask-claude.py "Reply with OK"). A failure here is a credentials problem to fix now, not during the first measurement run.

4. Initialize state

  • Rewrite memory/state.md: setup complete, panel empty, next steps /target-prompts-discovery and a first /site-readiness run (the day-one quick win: it needs no panel and its findings are usually cheap fixes).
  • Append the setup session to memory/journal.md.

Re-running

Safe to re-run any time facts change (rebrand, new domain, engine change, integration added). It updates config.json and the company profile; it never touches the panel, snapshots, or content items.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.