LLM Visibility Agent
On-demand editorial intelligence layer on top of the Amplitude or PostHog AI Visibility MCP. Pulls the latest weekly visibility report for your tracked org brand, compares it to prior weeks for trend, scores every gap where a tracked competitor shows up in an LLM answer but your brand does not, and produces a ranked list of recommended posts. Surfaces model-specific gaps (strong in Claude, absent in Google AI Overview), source and citation gaps (which domains the LLMs cite instead), and writes a summary brief to a path you confirm. Pure read and synthesize. Does not crawl LLMs, scrape, or call any model directly. Built for uristocrat.com and generalized so any brand tracked in Amplitude or PostHog AI Visibility can use it.
MCP-powered — requires setup
This skill uses external tools via the Model Context Protocol. You'll need to configure the following MCPs before installing.
Required MCPs
Reads weekly AI Visibility reports: scores, topics, prompts, competitors, sources, and per-model rankings for your configured org brand
Setup steps
- Set up each required MCP using the configs above
- Download the skill file below
- Open the skill manager in Claude — in the desktop app underCode → Customize, or on claude.ai underCustomize → Skills
- Click Create a new skill (use + on claude.ai) and upload the downloaded file
- Start a new session — your MCPs and skill will both be active
- Use a trigger phrase to activate
The MCP setup steps above are written for Claude. If your platform also supports MCP, configure the same servers using its own MCP settings. The skill file works either way.
- Download the skill file below
- Place it at
.cursor/skills/llm-visibility-agent/SKILL.md(rename the downloaded file toSKILL.md). The shared.agents/skills/llm-visibility-agent/SKILL.mdpath also works - Cursor loads the skill when your task matches its description. Say a trigger phrase to activate it
The MCP setup steps above are written for Claude. If your platform also supports MCP, configure the same servers using its own MCP settings. The skill file works either way.
- Download the skill file below
- In ChatGPT, create a Project and attach the file (or paste its contents into the project instructions). For a reusable setup, build a Custom GPT via Explore GPTs → Create and add the file there
- Start a chat in that Project or GPT, then say a trigger phrase to activate it
The MCP setup steps above are written for Claude. If your platform also supports MCP, configure the same servers using its own MCP settings. The skill file works either way.
- Download the skill file below
- Place it at
.github/skills/llm-visibility-agent/SKILL.mdin your repository (rename the downloaded file toSKILL.md) - Copilot loads the skill when it is relevant to your task. Say a trigger phrase to activate it
The MCP setup steps above are written for Claude. If your platform also supports MCP, configure the same servers using its own MCP settings. The skill file works either way.
- Download the skill file below
- Place it at
.agents/skills/llm-visibility-agent/SKILL.mdin your repository (rename the downloaded file toSKILL.md) - Codex loads the skill when your task matches, or invoke it explicitly with
/skills. Say a trigger phrase to activate it
Your skill is downloading.
Get an email when we ship new skills and updates.
You're on the list. We'll email you when new skills drop.
What this skill does
This is an overview of the skill from the site, not the skill's real instructions. The full SKILL.md Claude reads when this skill is active lives on GitHub at the pinned source below.
LLM Visibility Agent
Role
You are the editorial intelligence layer on top of a brand’s AI Visibility data. The measurement engine already runs weekly via the Amplitude or PostHog AI Visibility MCP. Your job is to read the latest report, compare it to prior weeks, find the prompts where competitors win and your brand loses, and hand back 3 to 5 concrete editorial moves.
You do not crawl LLMs. You do not call models. Everything you cite comes from the MCP tools.
Built for uristocrat.com and generalized so any brand tracked in Amplitude or PostHog AI Visibility can use it.
When to Activate
When the user asks any version of:
- “check AI visibility”
- “how is my brand showing up in LLMs”
- “run the LLM visibility report”
- “what AI prompts is my brand missing from”
- “AI search visibility brief”
- “am I showing up in AI search”
- “LLM visibility audit”
- “what should I write to rank in ChatGPT”
- “why is my brand not in AI answers”
- “/llm-visibility”
Step 0: Confirm
Before any analysis, confirm:
- The Amplitude or PostHog AI Visibility MCP is connected (the tools starting with
list_ai_visibility_org_brandsare available). If not, stop and tell the user how to connect it. - Which brand to analyze. Call
list_ai_visibility_org_brands, show the brands, ask the user to pick. If only one is returned, confirm by name. - Where to write the summary brief. Default suggestion:
~/notes/llm-visibility/YYYY-MM-DD-brief.md. If the user says skip, output inline only.
Step 1: Pull the Latest Report
Use the confirmed orgBrandId. Call get_ai_visibility_reports and pick the latest with status: completed. In parallel, call: scores, scores_over_time, topics, prompts, competitors, prompt_responses, sources, sentiment, models, pages. State brand, report date, ID, models covered, and any failed models at the top of every output.
Step 2: Trend
Compare the latest report against prior weeks per topic. Flag improving, flat, or declining on visibility score, average rank, and share of voice. Explicitly call out FAILED report weeks so trend lines are not misread.
Step 3: Gap Analysis
Per topic, list prompts where the brand visibility is 0 OR ranked below 3 AND a tracked competitor is mentioned. Score with gap_score = prompt_relevancy * competitor_strength * model_coverage_weight. Surface model-specific gaps and source citation gaps (which domains the LLMs cite when the brand is absent).
Step 4: Recommend
3 to 5 ranked recommendations. Each: topic, the absent prompt or prompts, a specific post idea, the competitor currently owning the answer, the model gap, the expected effect.
Hand-off block at the bottom in markdown list form so a downstream editorial or research skill can consume directly.
Step 5: Persist
Write to the path the user confirmed in Step 0. Filename uses report date: YYYY-MM-DD-brief.md. Frontmatter: brand, report_id, report_date, blended_score, topics_improving, topics_declining, gap_count. Link competitors as markdown links to their sites.
Output Rules
No em dashes. No filler. Numbers come from the MCP, not memory. The persisted file is summary only.
View full SKILL.md on GitHub