MCP analyst skills
The ten analyst skills bundled with GeckoGrid: what each one teaches an AI agent, and when the agent should load it.
Updated
GeckoGrid ships two kinds of skill markdown for AI agents. The operational skill (SKILL.md, served by get_mcp_skill_markdown) teaches an agent how to operate the app: the endpoint, the object shapes, recommended workflows, and the daily-scan guardrail. Agents load it right after connecting.
The analyst pack is different: ten skills that teach an agent how to interpret the data, the way a local SEO consultant would. They cover reading a scan, sizing up competitors, judging trends, and building both sales pitches and client reports. This page lists all ten.
How agents load them
The skills are compiled into the app, so they ship with it and work offline. A connected agent loads one over MCP with get_mcp_analyst_skill(name), passing the id shown below, and receives the full skill as markdown. The AI & MCP page lists the same skills with downloads, for offline or manual use, so you can also drop one into a workspace yourself.
The ten analyst skills
Rank diagnostics (rank-diagnostics)
Teaches the agent to read one completed scan: bucket the grid points by rank pattern, classify the underlying problem as relevance, proximity, or authority, and turn that diagnosis into specific recommendations instead of a summary of numbers. Load it when you ask “why am I not showing up” or “what does this scan mean”.
Competitor analysis (competitor-analysis)
Teaches the agent to aggregate the competitors captured at every grid point into a ranked list: who is outranking the business, where geographically, and what authority signals like rating and review count explain their lead. Load it for “who are my competitors” or “who ranks above me”.
Trend analysis (trend-analysis)
Teaches the agent to compare past scans of the same keyword and separate a real trajectory, improving, declining, or flat, from daily noise, including territory gained or lost and sudden drops. A meaningful trend needs at least three scans, and the skill says so. Load it for before and after questions and “is it getting better”.
Keyword research (keyword-research)
Teaches the agent to audit what a project already tracks, reason about the business’s actual services and market, and recommend keyword additions ranked by value for that specific business rather than a generic list. Load it when you want to expand tracking or ask “what keywords should I target”.
GBP audit (gbp-audit)
Teaches a complete Google Business Profile audit: categories, photos, posts, description, services, attributes, hours, and reviews, weighted by how much each factor moves local rankings, with an adaptation for service-area businesses that hide their address. Load it for “is my Google Business Profile complete”.
Website audit (website-audit)
Teaches an on-page local SEO audit of the website: title tags, meta descriptions, headings, LocalBusiness schema, NAP consistency, mobile and HTTPS, page structure, content gaps, and speed signals, ending in a prioritized fix list. It is designed to flow after the GBP audit, which surfaces the website URL. Load it for “audit my website”.
Schedule planning (schedule-planning)
Teaches the agent to pick a recurring scan cadence that lands inside the business’s open hours, avoids piling onto your existing schedules, and respects the daily-scan guardrail, then create the schedule. Load it for “when should I scan” or “set up recurring scans”.
Prospect report (prospect-report)
The writing methodology for the nine AI fields of a prospect report: headline, summary, highlights, design critique, fix plan, close pitch, and the email, DM, and walk-in outreach variants. The six client-facing fields appear in the report; the three outreach drafts are for you. It covers tone and the hard rules, above all never inventing a number that is not in the report’s own data. Load it when producing or improving a prospect report.
Prospect pipeline (prospect-pipeline)
The orchestration skill that takes a business from a name, an address, and one keyword to a finished, enriched report: creating the prospect project, running and waiting out the scan, matching the business in the results, adding market demand and money figures, comparing follow-ups, and exporting. It pairs with prospect report, which does the writing. Load it when the business is not tracked yet, or a pitch needs a refresh.
Client report (client-report)
The skill for reporting to a paying client: ask for your brand once and reuse it, pull the period’s numbers, export before and after map images, summarize competitor movement, optionally add a profile and website health check, and write the report in your format with the rank definitions named. Load it when you want to report results for a client GeckoGrid already tracks.
Was this helpful?
