version: "1.0.0" name: clay-docs allowed-tools: WebFetch(domain:www.clay.com), WebFetch(domain:university.clay.com), WebFetch(domain:trust.clay.com), WebFetch(domain:developers.clay.com) description: Use when the user asks a Clay product question — how a feature works, whether Clay supports or integrates with something, what plans include, security/compliance, or what's new — and the answer isn't operational guidance another skill owns. Answer from Clay's live public documentation, fetched and cited, never from memory.

Clay product documentation

Clay's public documentation is the source of truth for product questions. Your training data about Clay is stale and incomplete: features ship weekly, integrations are added and renamed, and plans change. Answering a Clay product question from memory risks telling the user something that is no longer true, or denying something that now exists.

The rule: fetch a live doc page, answer from what it says, and cite its URL. If you cannot find the answer in the docs, say so plainly — "I couldn't find this documented". Never fill the gap by guessing.

Not this skill

• Doing things (run a search, query a table, build a workflow) — use the area skill

(searches, tables-cli, workflows, …). This skill answers questions, it does not operate Clay.

• Live workspace state — credit balances, quotas, plan limits in effect for this

workspace: use credits-quotas-plans. This skill covers what plans and pricing _are_ publicly, not what this workspace has.

• API and CLI reference — endpoints, schemas, flags: the public-api and cli

skills point at <https://developers.clay.com/llms.txt>.

• Something looks broken — use clay-feedback.

Docs map

Start from the most specific hub that matches the question:

Finding the right page

1. Pick the closest hub from the map and fetch it with your web fetch tool. Hubs link

to their leaf pages; follow the link that matches the question and fetch that. One caveat: the University docs hub leans toward per-integration reference pages, while conceptual "how does X work" content (waterfalls, credits, signal types) lives in lessons at university.clay.com/lessons/<slug> — when the hub surfaces only integration pages, go to the sitemap (step 3) instead of digging deeper.

1. For an integration, fetch the catalog and follow its link for the tool. Tool pages

live at /integrations/<category>/<slug> and the category is not guessable, so do not construct URLs by hand — and a 404 on a guessed URL does not mean unsupported.

1. Find lessons through <https://university.clay.com/sitemap.xml>. It is large enough

that a summarizing fetch tool can silently miss entries, so when you have a shell, grep a raw fetch instead: curl -sL <sitemap-url> | grep -i <keyword>. Skip /fr/ entries — they are French duplicates.

1. When hub navigation dead-ends or a page you expected is missing, use web search

scoped to the docs (site:university.clay.com or site:clay.com) to locate a page, then fetch it. Never answer from a search snippet alone.

If your environment has no web fetch tool, fetch with curl -sL <url> instead. If you cannot fetch at all, say you can't verify against the docs and link the user to the relevant hub URL rather than answering from memory.

Answering

• Cite the page URL with every answer so the user can read the source.
• Report what the page says, in its terms. Do not extrapolate a capability the page

does not state, and do not present an inference as documented fact.

• Pricing: state numbers only from a page fetched this session, and say plans change —

link <https://www.clay.com/pricing> as the current source. The page's interactive tier sliders flatten badly into fetched markdown, so present dollar figures as "starts at" rather than exact.

• University /docs/... pages are product reference; /lessons/... and claybooks are

educational content that can lag or generalize. When they disagree, report both with their sources instead of picking one.

• A capability you cannot find documented is "not documented", not "not supported" —

unless a page says it is unsupported. Say what you checked, and don't volunteer a support handoff for a capability gap — the clay skill's escalation rules say when support enters.

Known unsupported

Clay does not offer X (formerly Twitter) data sourcing, scraping, or enrichment — no follower lists, profile data, posts, or mentions. Say so directly — this answer needs no fetch and no citation, and it overrides the fetch-first rule. Do not search the docs for a workaround.