Sentient's EvoSkill Now Runs Self-Improving Coding Agents Without Claude API
EvoSkill v1.3.0 adds Fireworks AI as a first-class inference provider, letting you run fully open-source self-improvement loops at high speed and low cost
- EvoSkill v1.3.0 adds Fireworks AI as a first-class provider for both the agent harness and the LLM scorer.
- A single
FIREWORKS_API_KEYenv var is all that's needed; EvoSkill auto-mirrors it across all internal consumers. - EvoSkill automatically evolves coding agents by analyzing failures and proposing skill/prompt mutations each iteration.
- Benchmarks show +7.3% on OfficeQA and +12.1% on SealQA, with zero-shot skill transfer to BrowseComp (+5.3%).
- Fireworks processes 13T+ tokens/day and leads GPU-based providers on latency, making it well-suited for high-iteration evolution loops.
- No breaking changes; existing Anthropic, OpenAI, and OpenRouter setups are unaffected. Goose requires manual OpenAI-compatible config.
EvoSkill, Sentient's open-source framework for automatically evolving coding agents, just shipped v1.3.0 with Fireworks AI as a fully supported inference provider. That means you can now run your entire agent self-improvement loop, both the evolution harness and the LLM scorer that judges results, entirely on fast, cheap open-source models. No Claude API required.
What EvoSkill actually does
EvoSkill is an open-source framework that automatically generates structured skills for AI agents by analyzing their failure cases, enabling continuous self-improvement. The core idea is to treat your agent's configuration, its system prompt and learned skills, as something that can be evolved rather than hand-tuned.
The self-improvement loop runs five stages on repeat:
- Base Agent attempts benchmark questions using the current best program.
- Proposer analyzes failures and suggests targeted skill or prompt changes.
- Generator writes the actual new skill files or rewrites the system prompt.
- Evaluator scores the new variant on a held-out validation set.
- Frontier tracks the top-N performing programs as git branches; the best survive to the next iteration.
EvoSkill significantly extends the feedback-driven idea of GEPA from single-file optimization to complete agent evolution. Instead of only revising one prompt in place, EvoSkill proposes multiple skill and prompt mutations jointly, evaluates new variants on held-out data, and has each iteration produce an entirely new agent program.