Bynn Agent Skills

Official Agent Skills for Bynn, Agemin, and Detector24. They teach any coding agent how to integrate document fraud detection for manipulated, fake and AI-generated PDF and image documents, plus identity verification, age verification, and content detection, including the parts that are easy to get subtly and dangerously wrong.

Works with Claude Code, Cursor, Codex, Copilot, Gemini CLI, OpenCode, Windsurf, Zed, and every other agent the skills CLI supports.

Onboard an agent in one shot

Point any coding agent at agent-setup.md and it will install the skills and wire up the MCP server itself:

Follow https://github.com/Bynn-Intelligence/skills/blob/main/agent-setup.md

It covers Claude Code, Cursor, VS Code, Windsurf, Zed, Claude Desktop, ChatGPT, and headless or CI setups, and tells the agent what to verify before it reports success.

Install

Any agent:

npx skills add Bynn-Intelligence/skills

The installer shows a picker grouped by product (Bynn, Agemin, Detector24), so you can toggle a whole product with one keypress on its group header. Press a to select everything. Nothing is pre-selected, which is the CLI's behaviour and not configurable per repository.

To skip the picker entirely:

# every skill, every detected agent, no prompts
npx skills add Bynn-Intelligence/skills --all

# every skill, but choose the agents yourself
npx skills add Bynn-Intelligence/skills --skill '*'

# just one product's skills
npx skills add Bynn-Intelligence/skills --skill bynn --skill bynn-mcp --skill bynn-kyc-sessions

Running inside a coding agent skips the picker on its own and installs everything.

Claude Code, with the bundled MCP server:

/plugin marketplace add Bynn-Intelligence/skills
/plugin install bynn@bynn

What is in here

bynn

The identity platform. Start with the bynn skill: it routes to the rest.

Skill	Covers
bynn	Router. Auth model, API base, errors, conventions.
bynn-mcp	Connecting an agent to mcp.bynn.com and driving Bynn through tools.
bynn-document-fraud	Forgery, tampering, and AI-generation analysis of documents.
bynn-kyc-sessions	Identity verification sessions: liveness, NFC chip reads, preflight.
bynn-web-sdk	Embedding verification in a website with @bynn-intelligence/websdk.
bynn-age-verification	Age estimation and minor detection, selfie and liveness.
bynn-moderation	Every detection model over image, video, text, and audio.
bynn-face-search	Face collections, enrollment, and 1:N search.
bynn-autodoc	Document requests sent over email or SMS.
bynn-webhooks	Payloads, the event catalogue, and signature verification.

agemin

Biometric age verification for websites and apps.

Skill	Covers
agemin	Router. Asset ID versus private key, and why the browser result is not proof.
agemin-web-sdk	@bynn-intelligence/agemin-sdk, age gating, SEO bypass, React and Vue.
agemin-api	Server-side confirmation, email and selfie checks, reference lookups.

detector24

Detection models for image, video, audio, and text.

Skill	Covers
detector24	Reading a model's input_schema and output_schema, async submission, and the rule that media must never be preprocessed.

Credentials

Nothing here ships a credential. Each skill says which key it needs and where to get it.

Product	Get keys at
Bynn	https://dashboard.bynn.com (agent token at /authenticate)
Agemin	https://agemin.com/app
Detector24	https://detector24.ai/app

Connect the MCP server

Installing skills does not connect the Bynn MCP server: npx skills add copies skill files only. Connect it separately to get the Bynn tools.

claude mcp add --transport http bynn https://mcp.bynn.com

Then run /mcp and sign in. The Claude Code plugin registers this URL for you, but it stays inert until that sign-in completes. For headless use, pass a token instead:

claude mcp add --transport http bynn https://mcp.bynn.com \
  --header "Authorization: Bearer $BYNN_API_TOKEN"

The endpoint is the root path. Appending /mcp returns a 404. Other clients, including Claude Desktop, ChatGPT, Cursor, VS Code, Windsurf, and Zed, are covered in the bynn-mcp skill.

Two rules worth reading before you build

Never preprocess media. Do not resize, crop, re-encode, or strip metadata before submitting an image, video, or document for fraud, tampering, or AI-generation analysis. Those detectors read artifacts that live in the exact bytes of the original file, and processing destroys them silently. The result comes back looking clean.

The browser is never the verdict. SDK callbacks run on the visitor's own machine. Confirm every result server side before unlocking anything.

Documentation

• Agent onboarding: agent-setup.md
• Bynn: https://docs.bynn.com, spec at https://api.bynn.com/openapi.json
• Agemin: https://agemin.com/docs
• Detector24: https://detector24.ai/docs

License

MIT. See LICENSE.