skf-create-skill

You can compile verified agent skills from skill-brief.yaml files using AST-backed structural extraction and provenance tracking. This process ensures zero-hallucination tolerance by validating extracted APIs against entry points and generating agentskills.io-compliant output.

The best way to extract a public API for skill compilation is using AST-backed structural extraction via ast-grep with semantic pre-ranking. This automatically validates the extracted API against the entry point and produces a compliant SKILL.md file with references and context snippets.

Yes, you can batch process multiple skill-brief.yaml files from a directory using --batch mode. This batch compilation includes checkpoint and resume capabilities, allowing you to pause and continue the extraction workflow without losing progress.

Provenance tracking during AST structural extraction works by generating complete provenance maps with line-level citations and confidence tiers. It creates evidence reports for full auditability, ensuring every piece of extracted code is traceable to its exact source location.

Yes, skill compilation supports Quick, Forge, Forge+, and Deep extraction tiers. These tiers apply AST-backed structural extraction and CCC semantic pre-ranking to handle varying codebase complexities and software engineering workflow requirements.

You need agentskills.io-compliant output to ensure your compiled agent skills are compatible across multiple IDEs and platforms. It enforces zero-hallucination tolerance by standardizing metadata, references, and context snippets into a validated SKILL.md format.