Researchers have introduced Code2Skill, a fully automated pipeline that extracts verifiable procedural skills directly from source code, according to a paper published on Hugging Face. The work addresses a key bottleneck in agent development: while reusable skills give agents transferable procedural knowledge, existing synthesis methods either require interaction with specific environments or rely on documents that may lack executable evidence.
Source code offers a complementary path, the authors argue, because it requires no prior agent experience yet provides executable evidence for grounding abstractions. Code2Skill transforms selected code units into implementation-anchored records of atomic operations, composite workflows, and recurring patterns, then verifies each record through source-body-blind reconstruction and source-aware comparison.
Applied to 19,769 popular, actively maintained GitHub repositories, the pipeline produced CodeSkillBank, a grounded bank of 1,006,822 accepted records with workflow, boundary, provenance, and source-evidence metadata. Across 72 protocol-matched evaluations covering nine model settings and eight benchmarks, models augmented with retrieved CodeSkillBank skills improved by 11.7% on average over matched baselines and outperformed them in 57 cases.
Under a unified downstream interface, Code2Skill also outperformed trajectory-derived skill banks on all seven shared benchmarks, suggesting that repository-derived skills can provide useful procedural knowledge before agents accumulate sufficient interaction experience. Skills synthesized from tested AI-generated code achieved a 93.50% pass rate, compared with 93.00% for human-written code, offering initial evidence that the pipeline can scale with the growing volume of AI-generated software.
The dataset is available at ant-intl/DeveloperSkills-Code2Skill, and CodeSkillBank powers the accompanying DeveloperSkillHubs website. Overall, the authors say, Code2Skill transforms procedural knowledge embedded in repositories into grounded, verifiable, and transferable agent skills.