Skrun vs Terrarium

Skrun is an open-source tool that wraps agentic skills — the discrete, reusable capabilities you build for AI agents (web search, data extraction, file transformation, API calls) — into deployable REST APIs with a single command. The idea is that skills you build for one agent context shouldn't be locked to that agent's runtime. With Skrun, you define a skill once with a standard function signature, and get a hosted endpoint with automatic request validation, retry logic, rate limiting, and an OpenAPI spec generated automatically. The project addresses a real architectural tension in the current AI tools ecosystem: agent skills are written in a dozen different formats (LangChain tools, MCP tools, function call JSON, OpenAI tool specs) and are essentially stranded assets — they only work within their specific orchestration framework. Skrun normalizes this by wrapping any skill definition format and exposing it as a framework-agnostic HTTP endpoint that any agent or pipeline can call. This appeared on Hacker News with a small but thoughtful discussion focused on the "skills as microservices" architectural pattern. Critics noted that adding HTTP round-trips to every tool call introduces latency; proponents argued that the composability and reusability benefits outweigh the cost. The early version focuses on stateless skills; stateful/conversational skill deployment is on the roadmap.

Terrarium is a multi-turn evaluation and optimization engine for LLM agents built by evolvent-ai. Unlike static benchmark suites that measure agents against fixed input-output pairs, Terrarium creates persistent, stateful "living environments" — simulated deployment contexts where agents operate over extended sessions, accumulate state, use tools, and interact with simulated external systems. You evaluate agents the way you'd test a car: by driving it, not by measuring its doors. The system supports configurable environment complexity, including simulated databases, APIs, file systems, and user personas. Agents are scored not just on final outputs but on trajectory quality — how efficiently they reached the answer, how often they hallucinated intermediate steps, and how well they recovered from dead ends. The engine also supports continuous optimization loops where poor-performing trajectories trigger automatic prompt refinement. With 17 stars and created April 14, Terrarium is extremely new. But it's addressing a genuine gap: the disconnect between how agents perform on static benchmarks versus how they behave in production. As enterprise AI deployments scale, the need for realistic pre-production evaluation is becoming critical.