Agent skill · security · aradotso
crabtrap-llm-proxy
LLM-as-a-judge HTTP/HTTPS proxy that secures AI agents by intercepting and evaluating outbound requests against security policies before they reach external APIs.
Why this skill is useful
Adds a comprehensive proxy setup with security policies, LLM evaluation, and PostgreSQL logging that fundamentally enhances agent security capabilities.
What it needs
Requires docker installed locally. Requires openai account access. About 7k tokens when loaded. Last updated 2026-07-09. 70 stars on the source repository.
What this skill does
CrabTrap LLM Proxy Skill by ara.so — Daily 2026 Skills collection. CrabTrap is a transparent HTTP/HTTPS forward proxy that sits between AI agents and external APIs. Every outbound request is intercepted, checked against deterministic static rules, then evaluated by an LLM judge against a natural-language security policy. Blocked requests return a 403 with a reason; all decisions are logged to PostgreSQL. Architecture Overview Key concepts: Static rules — deterministic prefix/exact/glob URL matching, checked first (no LLM call) LLM judge — natural-language policy evaluated only when no static rule matches Audit log — every request, decision, and response stored in PostgreSQL SSRF protection — blocks RFC 1918, loopback, link-local, and other private ranges Installation Docker Compose (Recommended) Initial Setup Configuration Full Configuration Reference Environment Variables CLI Commands Admin API All admin endpoints require Authorization: Bearer <admintoken>. User Management Static Rules Pattern types: prefix — URL must start with the pattern exact — URL must match exactly glob — wildcard matching with Rule priority: deny rules always take priority over allow rules. LLM Policies Audit Log Connecting an Agent Python Agent Example Using Environment Variables (Standard Proxy) Node.js Agent Example LangChain / OpenAI SDK Writing Effective Policies Policies are natural-language strings evaluated by the LLM judge. Be explicit about allowed and denied behaviors. Policy writing tips: 1. Start with what the agent is supposed to do (scope context for the LLM) 2. List explicit ALLOWED patterns before DENIED 3. Add a catch-all denial at the end 4. Deny rules in static rules are evaluated before LLM — use them for hard limits Policy Builder (Agentic Policy Generation) CrabTrap can analyze observed traffic and draft a policy automatically: The builder runs an agentic loop, analyzes recent audit entries, and proposes a policy draft for review in the UI. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills aradotso/crabtrap-llm-proxy