bp-skill-authoring

To write Agent Skill metadata and structure, define the YAML frontmatter with clear names, descriptions, and trigger conditions to ensure correct model recognition. Follow markdown guidelines to maintain concise, standardized content for accurate task triggering.

Progressive disclosure in skill authoring is a best practice that breaks long content into a main SKILL.md and on-demand loaded reference materials. This modular approach improves maintainability and ensures the model loads only necessary context for specific tasks.

YAML frontmatter requirements for a SKILL.md file specify standardized naming conventions for the skill name, description, and trigger conditions. These fields enforce community standards and ensure the skill is discoverable and correctly triggered by the agent.

Improve an existing SKILL.md structure by applying modular references, standardizing terminology, and using templated outputs. This structured refinement enhances the skill's maintainability and discoverability during future updates and model interactions.

Safety checks needed when creating an Agent Skill include evaluating dependency exposure and mitigating prompt-injection risks. These checks are enforced during the skill authoring process to ensure secure and reliable model operations and prevent unauthorized actions.

Move content into reference materials instead of the main SKILL.md when the content is extensive and only needed for specific tasks. Keeping the main file concise while using modular references allows the model to load expanded details on demand without exceeding context limits.