Innovate with AI. Securely.
Drop-in security recipes that embed Snyk directly into the AI coding assistants your developers already use — so AI-generated code is secure at inception, not patched after the fact.
Yes, if any of these apply:
- Your developers use Cursor, Claude Code, GitHub Copilot, Gemini, Windsurf, or Kiro — and you want every line of AI-assisted code scanned, fixed, and governed automatically.
- You're already a Snyk customer and want a fast, concrete way to apply Secure at Inception to agentic development without writing custom integrations.
- You're evaluating how to keep AI velocity from outrunning your security posture.
If you're new to AI-assisted ("agentic") development or haven't heard the term Secure at Inception, the What is Secure at Inception? section below is the 60-second primer.
A curated catalog of recipes — security building blocks that snap into your AI coding assistant:
| Recipe type | What it does | When it fires |
|---|---|---|
| Secure at Inception guardrails | Auto-scan AI-generated code for vulnerabilities the moment it's written. | Continuously, as the assistant writes code. |
| Secure at Commit guardrails (experimental) | Block git commit when staged code or dependencies introduce new vulnerabilities — a deterministic gate that fires no matter how the code was written. |
At git commit time. |
| Package enforcement guardrails | Block risky or vulnerable dependencies before they're installed. | At dependency-add time. |
| Remediation commands & skills | Find, fix, validate, and PR security issues in one guided flow (/snyk-fix, /snyk-batch-fix, secure dependency health checks, and more). |
On demand from the developer or the agent. |
| Snyk MCP integration | Wires Snyk's CLI and scanners into the assistant via the Snyk MCP server. | Automatically, once installed. |
All of it ships as one self-extracting installer. One command, every supported assistant configured.
AI coding assistants generate code at machine speed. Traditional security tooling — scanning at PR time, in CI, or at deploy — runs after the code is written and reviewed, which means risky code has already shaped the design, the tests, and the developer's mental model by the time anyone sees a finding.
Secure at Inception flips that: security checks run at the moment of generation, inside the assistant's loop, so vulnerabilities are caught and fixed before they ever land in a commit. It's Snyk's approach to keeping AI-driven development fast and safe — preventing new AI-generated risk while remediating existing security debt at the same speed your team is now writing code.
These recipes are the implementation. Snyk Studio is the platform.
1. Prerequisites
- A supported AI coding assistant (Cursor, Claude Code, GitHub Copilot CLI or VS Code, Gemini, Windsurf, or Kiro)
- A Snyk account — once installed, run
snyk author setSNYK_TOKENso the recipes can scan
The installer takes care of the rest: it bootstraps uv (Python), Node.js, npm, and the Snyk CLI for you, prompting before each step.
2. Install
Download and run the latest pre-built installer from snyk/studio-recipes:
macOS / Linux
curl -fsSL 'https://raw.githubusercontent.com/snyk/studio-recipes/main/installer/dist/snyk-studio-install.sh' -o snyk-studio-install.sh
bash ./snyk-studio-install.shWindows
powershell -Command "Invoke-WebRequest -Uri 'https://raw.githubusercontent.com/snyk/studio-recipes/main/installer/dist/snyk-studio-install.ps1' -OutFile snyk-studio-install.ps1"
powershell -ExecutionPolicy Bypass -File .\snyk-studio-install.ps1The installer auto-detects your assistants and merges in the recipes. Use --dry-run to preview, --ade <name> to target a single tool, or --uninstall to remove cleanly. Full options in installer/README.md.
3. Authenticate
Authenticate the Snyk CLI so the installed recipes can scan:
snyk authOr set the SNYK_TOKEN environment variable from your Snyk account for non-interactive environments (CI, containers, shared workstations).
4. Try it
Open your assistant and ask it to write code that touches a database, a network call, or a new dependency. You should see Snyk run automatically. Then try /snyk-fix on an existing project to see remediation in action.
I'm a developer or tech lead — start with the installer and the default profile. It gives you Secure at Inception guardrails plus on-demand fix commands across all detected assistants.
I'm a security professional or platform owner — review the recipe catalog under guardrail_directives/ and command_directives/ to understand which controls deploy, then standardize the default (or a custom) profile across your org via your usual provisioning channel.
I'm a technical decision maker evaluating Snyk — read the What is Secure at Inception? section, then read about Snyk Studio and the Snyk Platform. These recipes are the developer-facing surface of that platform.
I want fine-grained control — every recipe is a plain file under guardrail_directives/, command_directives/, and mcp/. Copy, fork, customize, and combine — they're designed to be layered.
Cursor • Claude Code • GitHub Copilot (CLI and VS Code) • Gemini • Windsurf • Kiro
Coverage varies slightly by tool (e.g. Copilot CLI has no custom slash commands yet, so it gets skills + MCP only). The installer README has the exact matrix.
These are starting points, not rigid templates. Common customizations:
- Adjust scan thresholds and severity gates
- Add or remove workflow phases in remediation skills
- Compose multiple recipes into a workflow tuned to your team
- Adapt patterns to assistants not yet covered out of the box
Snyk Studio is built on an open, partner-first, tool-agnostic foundation — these recipes follow the same philosophy.
- Snyk Studio product page — the product these recipes plug into
- Snyk Platform — the broader AI-native security platform
- Snyk Studio documentation — setup, MCP server, and reference material
guardrail_directives/— automatic, always-on controlscommand_directives/— on-demand security workflowsinstaller/README.md— installer flags, profiles, and verification
Reach out to your Snyk account team, or open an issue in this repository.
This repository is closed to public contributions.