Build with AI

Let your coding agent write the integration. The Enforly skill teaches it to check before the effect, send the right evidence in data, and stop on anything other than allow.

Install the skill

Run this in your project. It installs the skill for Claude Code, Cursor, Codex, GitHub Copilot, Windsurf, Cline, and other coding agents that support skills. You still need an API key from API keys in your server environment.

Terminal
npx skills add CodeSquar/enforly-skills

Or copy skills/enforly from the repository into your agent’s skills directory, for example .claude/skills/enforly.

Ask your agent

Describe what to guard and the rule in your own words. For example:

  • “Guard the tools of my agent with Enforly so it can’t drop tables without confirmation.”
  • “Check user posts with Enforly before publishing them.”
  • “Require a manager approval from the ticket before refunds above 200 USD.”

What the skill covers

Setup

Install @enforly/sdk and keep ENFORLY_API_KEY on the server.

Decisions

Continue only on allow. Stop on deny, review, and every error.

Evidence

Put the exact operation and the evidence the policy mentions in data.

Policies

Write rules that name their exception and who approves, and resolve a review by checking again with the approval.

Recipes

Agent tool calls, a Claude Code hook, and checks outside AI workflows.

Agents without skills

Point the agent at llms.txt, a plain-text summary of the same rules, and name what to guard.

Prompt
Read https://enforly.com/llms.txt and follow its rules to add an Enforly check
before <the operation to guard>. Only continue when the result is allowed;
stop on deny, review, and every error.
Writing it yourself? Open SDK docs or API docs