Analytics skills

Agent skills that teach Claude, Cursor, and other AI agents to read web analytics like a senior analyst. Diagnose traffic changes, judge channel quality, and read funnels without the usual rookie mistakes. Works with any analytics source; first-class support for Clamp MCP.

What you get

Five model-invoked skills following the open Agent Skills spec. Claude loads the right one based on the question you ask. On first run, analytics-profile-setup interviews you about your business and writes a profile file; every other skill calibrates its answers to your industry, scale, and ICP instead of reaching for generic averages.

Paste to your agent

Install the Clamp Analytics Skills

Adds the plugin marketplace, installs the skills pack, and runs the one-time profile setup so future answers are calibrated.

Install the Clamp Analytics Skills in this Claude Code project.

Read https://clamp.sh/docs/skills.

1. Run `/plugin marketplace add clamp-sh/analytics-skills` to add the marketplace.
2. Run `/plugin install analytics-skills@clamp-sh` to install the pack.
3. After installation, run `/analytics-skills:analytics-profile-setup` and answer the interview so every other skill calibrates to this business (model, primary conversion, traffic range, ICP, data stack). The profile lands in a local file other skills read.
4. Confirm the install worked by listing the available `/analytics-skills:*` commands. Do not invent skill names — they're listed at https://clamp.sh/docs/skills.

Install

Claude Code plugin (recommended)

terminal

/plugin marketplace add clamp-sh/analytics-skills
/plugin install analytics-skills@clamp-sh

Skills become available as /analytics-skills:<name> and Claude auto-invokes them when the task matches.

Claude Code standalone

Clone into your personal skills directory:

terminal

git clone https://github.com/clamp-sh/analytics-skills.git /tmp/analytics-skills
cp -R /tmp/analytics-skills/skills/* ~/.claude/skills/

claude.ai

Pre-built ZIPs are attached to every GitHub release. Upload each one via Settings → Features → Skills.

The skills

Skill What it does

analytics-profile-setup One-time interview that captures your business context (model, primary conversion, traffic range, ICP, data stack) into a local profile file. Run this first.

analytics-diagnostic-method The spine. Five-step method: load profile, frame the question, build a MECE hypothesis tree, triangulate, present with the Pyramid Principle. Other skills build on this.

traffic-change-diagnosis Drill path for “why did traffic change”. Fingerprints for tracking regressions, bot spikes, deploy-correlated drops, campaign ramps, SEO decay.

channel-and-funnel-quality Volume × engagement × conversion as a matrix. Vanity-traffic detection, expected drop-off ranges by funnel step, mix-shift handling.

metric-context-and-benchmarks What’s a good bounce, engagement, duration, CVR, churn, LTV:CAC, or activation by business model. When each metric lies and minimum sample sizes before trusting a rate.

Skill	What it does
`analytics-profile-setup`	One-time interview that captures your business context (model, primary conversion, traffic range, ICP, data stack) into a local profile file. Run this first.
`analytics-diagnostic-method`	The spine. Five-step method: load profile, frame the question, build a MECE hypothesis tree, triangulate, present with the Pyramid Principle. Other skills build on this.
`traffic-change-diagnosis`	Drill path for “why did traffic change”. Fingerprints for tracking regressions, bot spikes, deploy-correlated drops, campaign ramps, SEO decay.
`channel-and-funnel-quality`	Volume × engagement × conversion as a matrix. Vanity-traffic detection, expected drop-off ranges by funnel step, mix-shift handling.
`metric-context-and-benchmarks`	What’s a good bounce, engagement, duration, CVR, churn, LTV:CAC, or activation by business model. When each metric lies and minimum sample sizes before trusting a rate.

Works with Clamp MCP

The skills are analytics-agnostic but include Clamp-specific tool call-outs. With Clamp MCP connected, the agent knows exactly which tool to reach for at each step of the method (traffic.overview, referrers.top, funnels.create, etc.). Using a different analytics source? The method still applies; only the tool names change.

Source and changelog on GitHub. MIT licensed.