evolving-skill-creator

The Evolving Skill Creator is a specialized agentic workflow system designed to turn repetitive manual processes into robust, reusable skills. It encourages an iterative development cycle where tasks are captured as structured workflow plans, accompanied by purpose-built Python scripts that handle execution logic. This skill is ideal for developers, engineers, and researchers who find themselves performing the same sequences of data extraction, file processing, or system interaction, allowing them to codify 'best practices' into an evolving knowledge base that improves with each cycle.

• Enables the creation of machine-readable SKILL.md files using YAML frontmatter for seamless skill discovery.

• Promotes a 'plan-before-execute' methodology by forcing documentation of Python scripts, inputs, outputs, and dependencies before code creation.

• Supports systematic refinement through dedicated sections for capturing learnings, 'what worked,' and 'tips for future use' to ensure continuous improvement.

• Facilitates directory-level encapsulation of logic, ensuring that evolving skills remain modular, portable, and easily shareable across different agent environments.

• Utilize for tasks involving web scraping, static site generation, complex data transformation, or iterative research projects.

• Structure your workspace into tasks/evolving_skill/ directories, maintaining clear separation between workflow documentation (SKILL.md) and executable Python tools in scripts/.

• Always define clear tool purposes, inputs, and expected outputs in the Tools to Create section to aid agent understanding and future maintenance.

• Before execution, check if existing custom tools or MCP (Model Context Protocol) servers can reduce the need for custom Python code.

• After completion, update the Learnings section with post-execution insights to improve the skill for subsequent runs or for other team members.