Create and iteratively improve Herald skills with structured SKILL.md files.
npx skills add https://github.com/dbbaskette/herald --skill skill-creator-dbbaskette
Please help me install this Agent Skill. Skill: skill-creator Source: https://github.com/dbbaskette/herald/tree/main/skills/skill-creator Command: npx skills add https://github.com/dbbaskette/herald --skill skill-creator-dbbaskette
This meta-skill provides a structured workflow to create, test, and iteratively improve Herald skills, reducing cognitive load and enabling repeatable, measurable skill development from concept to deployment.
Define a new skill by creating a SKILL.md with a name and description, then add optional scripts, references, and assets to support evaluation and iterative improvement.
High-intent search queries and answers about installing and using this skill.
To create and iterate on skills, you define a SKILL.md file with YAML frontmatter containing a name and description, add actionable instructions in the body, and use an iterative evaluation workflow to test and improve the output.
Automating skill evaluation involves generating eval prompts, producing an evals.json file for automated grading, running controlled tests, and visualizing benchmarks to inform iterative improvements to the skill instructions.
Yes, you can optionally bundle scripts, references, and assets with your skill to support deterministic tasks, in-depth research, and output templates, ensuring repeatable results and reducing cognitive load.
No dependencies are required to generate evaluation prompts. The skill creation workflow operates independently, reading optional bundled scripts only when they are provided to support the core SKILL.md body.
A SKILL.md body and frontmatter are needed because the frontmatter provides the required name and description entry point, while the body delivers the actionable instructions and context necessary for the skill to execute correctly.
Structuring skill creation end-to-end involves defining a clear entry with frontmatter, writing actionable guidelines in the body, attaching optional bundles for repeatability, and running an iterative evaluation workflow to ensure measurable deployment.
nirukk52Create and package Claude skills with SKILL.md templates and validation.
matantsachRun existing skill evaluations via npx snapeval eval and parse benchmark results.
zhuxiningCreate modular Claude Skills with SKILL.md frontmatter and packaging scripts.
pbc-osIteratively improve measurable systems through hypothesis-driven single parameter changes.
