The Agent SkillThe four integration steps, the account model, the framework-specific traps, and how to verify. Ships in every pack and standalone, open source.
qid-connect-skill on GitHub
For the agents, and the humans who run them
Make your agent a BTX app builder
Hand any AI coding assistant one prompt and it knows qID: the post-quantum login, the account model where the address is the account, the framework traps that produce no error in the browser, and how to verify the result against your running app. Free of charge, MIT, no API keys, nothing to sign up for.
The prompt
Paste this into your agent. It works in Claude Code, Cursor, or anything that can fetch a URL.
Read https://qid.dev/llms.txt and follow its links. Then add Sign in with qID to my app: download https://qid.dev/connect/qid-connect-latest.zip, follow docs/INTEGRATION.md and skills/qid-connect/SKILL.md from the pack for my stack, and verify the result with skills/qid-connect/scripts/check-integration.mjs against my running dev server. Remember: the address is the account, a login proof can never move funds, and since 1.7.0 the QR poll returns status "confirm" before "done".
What your agent gets
The checkerZero dependencies, runs against your live integration, catches the failures that show nothing in the browser. Each failure prints its fix.
check-integration.mjs
check-integration.mjs
llms.txtThe whole estate, machine-readable: protocol facts, endpoints, the 1.7.0 confirm contract, every fetchable link.
qid.dev/llms.txt
qid.dev/llms.txt
The packServer SDK, widget, examples, docs and the skill in one MIT zip. What an integrator gets, whether human or agent.
qid-connect-latest.zip
qid-connect-latest.zip
Install the skill directly
For Claude Code, from a clone or the unzipped pack:
git clone https://github.com/MendeMatthias/qid-connect-skill mkdir -p .claude/skills && cp -R qid-connect-skill .claude/skills/qid-connect
For every other assistant: point it at the folder, or have it fetch SKILL.md and the references directly. The skill is plain markdown plus one script; nothing to build.
Why this works
- Everything is fetchable. Site, docs, pack, skill, checker, changelog, timeline: public URLs, no keys, no accounts.
- The integration is small. One server middleware plus one button. A developer lands a real post-quantum sign-in in 15 to 30 minutes; an agent with the skill does the same.
- Verification is built in. The checker proves the integration outside-in, so "it looks done" becomes "it is done".
- Safe by construction. Login proofs sign under a dedicated domain tag and can never move funds. Since 1.7.0, QR sign-in requires the user to confirm the address before any session exists.
- Built for what comes next. The same login works for any BTX app today and is designed to carry forward: post-quantum from day one, so the apps you build now do not need rescuing later.
The honest line, as always. Resistant to quantum attacks, built on NIST-standardized post-quantum cryptography, pending independent audit. Never "unhackable". The qID Connect pack and the skill are MIT and free; the core library opens with the audit. Questions from an agent's human: open an issue.