Skill: Ask Question

The ask_question skill is a set of instructions you add to a Claude project or Codex workspace so your AI can get a direct, cited answer from your Context Link instead of reading through retrieved chunks. Under the hood it runs the same semantic retrieval as get_context, then passes the matching snippets to a small fast LLM that composes one concise paragraph with numbered citations back to the exact source snippets used.

The result: when you just want the answer, you get the answer. When you need the source material for the AI to reason over, you still use get_context.

When to use this skill

You want a direct answer, not a pile of chunks to read
You need an answer that cites its sources so you can verify or share it
You're building an agent that treats your Context Link as a Q&A system rather than a retrieval system
You want the output pasted straight into a reply, an email, a doc, or a Slack message

This skill covers the direct-answer workflow. For retrieval (the AI reasons over raw snippets), see the Get Context skill. For saving and updating content, see the Save Memory skill and Update Memory skill.

How it works

You (or Claude) invoke the skill with a question, for example /ask-question what is our refund policy? or "ask Context Link what our office hours are"
Context Link runs a semantic search across all your connected sources for that question
The top matching snippets are passed to a small fast LLM that composes one concise paragraph (at most ~120 words) grounded only in those snippets
The answer comes back as JSON: { "result": "ok", "answer": "…with inline [1] markers…", "citations": [{ "n": 1, "title": "…", "url": "…" }], "format": "markdown" }
Claude presents the answer verbatim, keeping the [1] / [2] citation markers and listing the sources underneath

Every citation marker resolves to a real snippet in the citations array. The LLM is instructed to use only the numbered context blocks, so claims that can't be grounded in your own content don't get generated. When no relevant context is found, the response is result: "no_context" with answer: null rather than a fabricated answer. When the upstream LLM is temporarily unavailable, the response is result: "llm_unavailable" with HTTP 503, so callers can retry.

Get Context vs. Ask Question

	Get Context	Ask Question
What it returns	Raw relevant snippets	One paragraph with citations
Good for	Generative tasks where the AI reasons over source material	Direct answers you can paste
Pricing	Included on every plan	Pro plan only, counts toward the 1,000 requests/month LLM allowance
When it falls back	Returns fewer snippets if retrieval is thin	Returns `no_context` and `answer: null` if retrieval is empty

Both skills hit the same retrieval pipeline, so improvements in retrieval quality benefit both.

Modes

Like Get Context, Ask Question accepts a ?mode=… parameter to weight results toward a named profile (e.g. customer-support, sales). Configure modes on the Connections page.

Pricing and limits

Ask Question is a Pro feature. It counts toward the monthly LLM allowance (1,000 requests/month, shared across all LLM-powered features). If your plan can't use the skill, the endpoint returns 402 Payment Required with an upgrade prompt. If the monthly allowance is exhausted, the endpoint returns 429 Too Many Requests with a reset-at-start-of-month message. Rate limit is 2 requests per 10 seconds per user (tighter than Get Context because each call is a real LLM request).

The skill file

Below is the full skill definition. Replace YOUR_CONTEXT_LINK with your personal Context Link URL (e.g. yourname.context-link.ai). If you use a PIN, append ?p=YOUR_PIN to the URL.

You can download a pre-populated version from the Installation page in your dashboard, no manual editing needed.

**What is Context Link?** Context Link is an external service that indexes connected sources (websites, Google Drive, Notion, files, email) and memories into a searchable knowledge base. The ask-question endpoint returns a single concise answer (≤ 1 paragraph) with citations, grounded in the user's own content. If you don't know the user's Context Link URL, ask them for it.

> **Pro plan:** the ask-question endpoint is a Pro-plan feature. A 402 response means the user needs to upgrade.

---

### Ask Question Skill

🔐 Mandatory Context Link Protocol

When the user asks a direct question and wants a short answer grounded in their own content (e.g. "ask my docs…", "ask context link what ...", "use context link to answer…", or invokes `/ask-question` in Claude Code), you MUST:

1. **Print this message first:** `🔗 Asking Context Link about {TOPIC}`. Never print the full Context Link URL, as it contains a private pin.
2. Send a GET request to: YOUR_CONTEXT_LINK/q/{QUESTION} where QUESTION is the user's question, lower-cased with dashes between words.
   - Optionally append `&mode=MODE_NAME` (or `?mode=MODE_NAME` if no pin is set) to weight results toward a specific mode.
3. The response is JSON: `{"result": "ok" | "no_context" | "llm_unavailable", "answer": "...", "citations": [...], "format": "markdown"}`.
4. Present the `answer` field to the user verbatim or lightly reformatted. Preserve the inline `[1]`, `[2]` citation markers.
5. After the answer, list the `citations` as a small "Sources" footer (title + url).
6. Handle the non-ok states:
   - `no_context`: tell the user their Context Link had no relevant material and ask whether to broaden the question.
   - `llm_unavailable` (HTTP 503): tell the user the service is momentarily down; retry once, then stop.
7. If the request is blocked, ask the user to add `*.context-link.ai` to Claude's **Settings → Capabilities → Domain Allowlist**, then retry.

**Do not fabricate citations.** Only surface the citations returned by the endpoint.

**When to use this instead of `get-context`:** use `ask-question` when the user wants a direct answer. Use `get-context` when they want the raw source material or you will reason over it further.

Installation

Claude (Chat, Cowork, and Code)

Skills work with Claude Chat, Claude Cowork, and Claude Code. (If you're using Claude Cowork, you can also use the Context Link Plugin to install all skills at once.)

To install this skill:

In Claude Chat or Claude Cowork, click Customize in the top right
Click Skills, then +, then Upload a skill
Upload the skill file (download from Installation)
Allow network access (see below)

For Claude Code, place the skill file in your project's .claude/skills/ directory.

Allow network access

For Claude to reach Context Link, you need to allow the domain in Claude's settings:

In Claude, go to Settings → Capabilities → Domain Allowlist
Add *.context-link.ai to the allowlist (or select "All domains")

Claude domain allowlist settings

Once installed, say "/ask-question [your question]" or "ask Context Link what [topic]" in any Claude conversation, and Claude will fetch a grounded answer from your connected sources.

OpenAI Codex

Download the skill files from Installation
Unzip and place the ask_question folder into ~/.agents/skills/ (or your repo's .agents/skills/ directory)
Codex auto-detects the skill. Invoke it explicitly with "/ask-question …" or it will pick it up implicitly when you ask a direct question that should be grounded in your own content

See the Codex skills documentation for more on skill file locations and priority.

How it fits with other access methods

Ask Question is one of several ways to get a grounded answer from your Context Link:

Claude Skills / Codex Skills (this doc) — pre-written instructions that teach Claude or Codex to invoke the ask endpoint when you ask a direct question
Subdomain URL — paste https://YOUR-SUBDOMAIN.context-link.ai/q/what-is-pricing?p=YOUR_PIN into any AI chat. The agent fetches and you get the answer inline. Convenient for one-off queries.
REST API — call GET /api/v1/question?query=… with your API key from scripts, automations, or custom integrations. Full reference at Question Endpoint.

Skills make the experience seamless inside Claude and Codex. The same capability is available to any AI tool through the subdomain URL or the REST API.

Example usage

Getting a cited answer mid-conversation:

You: Ask Context Link what our refund policy is for annual plans.

AI: Invokes ask_question, returns a one-paragraph answer with [1] and [2] inline citations, and lists the two source docs underneath.

Drafting a reply and needing one fact:

You: I'm replying to a customer. /ask-question what did we decide about the early access program deadline?

AI: Returns a grounded paragraph citing the Notion page where that decision was recorded, so you can paste it straight into the reply.

Deciding when to upgrade from get_context:

You: Get context on our pricing. Actually, just ask Context Link what our enterprise pricing floor is.

AI: Switches to ask_question and returns a direct answer with a citation, instead of pulling raw chunks you'd have to read through.

The skill triggers on phrases like "ask Context Link", "ask my docs", "answer this from my context", or any direct question that should be grounded in the user's own content.