Agent skills for Claude Code and Cursor
Thousands of rows of raw events go in. A report comes out where every figure is tested against the file's own totals, and any figure it cannot prove is printed UNPROVEN instead of shipped. These skills teach your agent the craft.
Most buyers have never dropped one in. A skill is a folder of plain markdown your agent reads before it works. It turns a model from a clever first draft into a senior BI engineer that refuses to guess.
~/.claude/skills/ kymira-anchor-discovery/ SKILL.md references/ kymira-report-design/ SKILL.md kymira-verification-practice/ SKILL.md ... seventeen more
Copy the folder into ~/.claude/skills/. No build step, no runtime,
plain markdown you can read and edit.
Ask for the report you need, in plain language. "Did the promotion work." "What we shipped versus what actually sold." It builds against your real exports.
Every figure traces to its source cell. Anything the file cannot prove is named on the page, not shipped as fact.
A model on its own
A clever first draft, confident everywhere.
A model with the skills
A senior reviewer who shows the work.
Twelve teach the craft of a report you can trust. Seven make it a platform. One makes it yours, behind a wall, for your whole team.
✓ marks the tiers each skill ships in
✓ Everything · the superpower
Every report you build lands in one place on Cloudflare, behind a default-deny auth wall with per-audience zones. Refresh a dashboard by dropping in a new export. Shared team state updates live and persists, flags, boards and rulings, so the whole team works inside numbers that reconcile.
Ships with the four-layer architecture, the security-hardening gate, the CSP hashes and the backup drill you rehearse before anyone external logs in. The master intranet template is included.
See stage 3 of the quick start →These skills were hardened building a real reporting system, one pushed until it stopped shipping a single number it could not prove. Every rule in them exists because a real report once got a number wrong, and now it cannot.
The skills did not come from a blog post or a prompt pack. They were distilled from a real reporting system that was tested, broken on purpose, and rebuilt until it stopped producing wrong numbers. Every rule you get is one that a real failure paid for.
That makes them cautious the way a careful analyst is cautious. A figure either reconciles to the totals in the file it came from, or it is labeled UNPROVEN on the page. The skills would rather tell you what they cannot prove than hand you a confident wrong answer.
And nothing is taken on trust. Every number in a finished report traces straight back to the rows in the source file that produced it.
This number reached a live page once, published as if it were real. It is wildly wrong, and a single missing check is all it took to let it through. Catching this kind of error before anyone acts on it is what these skills are for.
One specimen · the Metric Library
How much recurring revenue this year's cohort of existing customers grew or shrank to, counting their upsells, downgrades, and cancellations, but not any brand new customers.
The fork, why two teams get different numbers
The single most gamed SaaS metric, and the forks live in the denominator and the window. The cohort fork: NRR must measure only the customers present at the start, yet the common error is ending total MRR over starting total MRR, which sneaks new customer revenue into a retention number. The window fork: a trailing twelve month cohort figure versus a single strong month annualized. Investors read the trailing figure, internal dashboards often show the annualized one, and they rarely match.
The trap
The denominator swap. A company adding logos fast can show 130% NRR while its actual cohort is shrinking, a growth rate wearing a retention label.
Every paid tier is a one-time purchase with lifetime updates. No seats, no subscription, no lock-in.
The files are plain text with no copy protection, because a tool you cannot read is not one you can trust. Own them once, edit them freely, keep them forever.
What is Kymira?
A set of installable agent skills and two reference libraries that teach an AI coding agent, Claude Code or Cursor, to turn raw data exports into business intelligence where every number reconciles to the file's own totals, and where anything unprovable is named rather than shipped.
What do the skills actually do?
Each skill is a plain markdown playbook your agent reads before it builds a report: how to find what a file asserts about itself and reconcile every figure to it, resolve the metric definitions teams disagree on, verify every figure against the source file, and design a dashboard that reads at a glance. Plus the skills that turn your reports into a secure internal platform, with maps, exports, delivery, alerts, decks, annotations, and the metric store. Twenty skills in all.
How do I install a skill?
Copy the skill's
folder into your agent's skills directory, ~/.claude/skills/ for Claude Code, or
the generated AGENTS.md rule for other agents, and start a fresh session. There is no build
step, a skill is just files. The install guide walks through each agent.
What is in the Metric and Format libraries?
The Metric Library has 346 metric definitions across twelve departments, each naming the competing definitions teams argue over (the fork) and how the number lies in real exports (the trap). The Format Library maps 47 vendor export shapes, Stripe, QuickBooks, Shopify, Salesforce, GA4 and more, with the structural traps in each and the anchor that reconciles it.
Do I own the skills, or is it a subscription?
You own them. Every paid tier is a one-time purchase with lifetime updates, no subscription. The files are plain text with no copy protection, because a tool you cannot read is not one you can trust.
Can you build it for me?
Yes. Built for you is a founder-led engagement: the engineer behind the engine builds a custom, certified BI site on your real exports. It is by application, priced on the scope.