Get started
Install the skills for the work in front of you, ask for that work in your own words, and check what comes back.
A skill is a folder of instructions your agent reads when your request matches it: a SKILL.md, sometimes a standard-library Python script, and the evals that check it. Installing one copies files you can read, diff and edit, and runs nothing. A skill that needs a command or a network host names it in its manifest.
Step 1: install
Needs Node 16.7 or later. Installs one pack into Claude Code's user folder; add --ide cursor, --ide copilot, --ide kiro or --ide codex for another tool.
Add --dry-run first to see every file it would write.
Add the catalogue once, then install the plugin for a pack:
A skill is a plain folder. Copying it is the whole install:
Step 2: choose your starting work
Each pack's first task is chosen so you can judge the result in minutes.
Dev team: Find out whether your current branch is safe to merge.
Then paste this into your agent:
Is this branch safe to merge? Run a pre-merge review on the current diff.
How to tell it worked: pre-merge-review runs the project's lint, typecheck and tests through its gate script, then the reviewer panel, and ends with READY or NOT READY.
Product manager: Pressure-test a product idea before anyone builds it. About 1–2 hours of working time, plus the interviews themselves.
Then paste this into your agent:
Write a PR/FAQ for a feature that lets customers pause their subscription for up to three months instead of cancelling.
How to tell it worked: prfaq returns a launch press release, a customer FAQ and an internal FAQ that names the hard questions.
Solution architect: Record an architecture decision your team already made. About 1–3 hours.
Then paste this into your agent:
Write an ADR: we're moving our session store from Redis to DynamoDB. We considered staying on Redis, DynamoDB, and Postgres; we chose DynamoDB for managed scaling and lower ops load.
How to tell it worked: adr-generator returns an ADR with context, the options considered with their trade-offs, the decision, and its consequences.
SRE / on-call: Write the runbook an on-call engineer can follow at 3am.
Then paste this into your agent:
Write a runbook for our payments-api service: it runs in the Kubernetes namespace payments, depends on Postgres and Stripe, and pages the payments on-call rotation.
How to tell it worked: runbook-generator returns a runbook with rollback commands, the common incidents and what to do for each, and an escalation path. Facts you did not give it are marked missing, not invented.
Data and analytics: Check a query before its number goes into a report.
Then paste this into your agent:
Review this Postgres query before it goes into the finance dashboard. One row per region; customers.customer_id is unique and an order has many order_items. Finance reports in New York time. SELECT c.region, SUM(o.amount_usd) AS revenue_usd FROM orders o JOIN customers c ON c.customer_id = o.customer_id JOIN order_items oi ON oi.order_id = o.order_id WHERE o.created_at BETWEEN '2026-09-01' AND '2026-09-30' GROUP BY c.region;
How to tell it worked: sql-review opens with a verdict, flags the order_items join as multiplying revenue and the BETWEEN as losing 30 September, gives a corrected query with New York boundaries, and lists verification queries to prove the fix.
Core: Turn rough notes into a structured brief, then get a verdict on the draft. About 20–45 minutes.
Then paste this into your agent:
Here are my notes from today's planning call: <paste them>. Turn them into a brief, then critique the draft you write from it.
How to tell it worked: brief-intake returns a structured brief with gaps tagged, and doc-critique ends with a verdict: SHIP IT, SHIP WITH CHANGES, MAJOR REWRITE or WRONG ARTIFACT.
Other packs, each with its own first task: Stakeholder comms · Design · AI engineering · API builder · Converters · Experience design · Research · Trackers · Governance, risk and compliance · Infrastructure as code · Frontend · Skill engineering · Job applications. Not sure which? Start with your role or compare two packs.
Step 3: make the first request
Ask in plain language. Your agent picks the skill from what you ask for, so there is no command to remember. You can also name a skill, or type /skill-name in Claude Code. Other first requests that work:
- I need to research whether our 14-person support team should move to a four-day week. Our COO decides by 1 November. I have about 20 hours and can pull ticket data from Zendesk.
- Build a risk register for our office relocation project. Risks: the lease on the new floor isn't signed yet (owner: Facilities Manager, Dana Cole), the network fit-out might slip past the move date (owner: IT Manager, Raj Patel), and staff may push back on desk sharing (owner: HR Business Partner, Lena Fox). Use a 5x5 scale; today is 2026-10-01.
- Convert reports/q3-review.md to a Word document for the Acme account team, and put its tables in an Excel workbook for finance.
Looking for a skill by what it does? npx skilldrop-cli search review my sql lists the best matches.