Create, test, evaluate, and refine Claude skills through a structured workflow.
npx skills add https://github.com/hackutd/harp --skill skill-creator-hackutd
Please help me install this Agent Skill. Skill: skill-creator Source: https://github.com/hackutd/harp/tree/main/.claude/skills/skill-creator Command: npx skills add https://github.com/hackutd/harp --skill skill-creator-hackutd
This Skill provides a structured process for designing, testing, evaluating, and improving Claude skills instead of relying on ad hoc instructions or subjective iteration.
Use the skill-creator skill to design or improve a Claude skill, create realistic evaluations, benchmark its performance, and optimize its triggering description.
High-intent search queries and answers about installing and using this skill.
To create and test Claude skills, use a structured workflow to capture intent, define trigger conditions, draft SKILL.md files, run baseline comparisons, and grade assertions for iterative improvement. This replaces ad hoc instructions with evaluated skill designs.
Evaluating prompt engineering skills requires creating test cases, running skill-enabled and baseline comparisons, grading assertions, and aggregating benchmark results to identify performance gaps and recurring weaknesses for iterative refinement.
Optimizing skill descriptions involves generating trigger evaluations to test activation accuracy, refining the descriptive metadata, and comparing successive versions to ensure the skill activates reliably under the intended conditions.
Yes, you can convert repeatable document-processing, coding, or research workflows into tested and discoverable Claude skills by defining triggering conditions, specifying output formats, and drafting complete SKILL.md files.
Inconsistent skill activation often stems from poorly optimized trigger descriptions. Running trigger evaluations and iterating on the activation description helps ensure the skill fires reliably for the intended workflows.
yu13130122297Identify uncertainties in research framing, datasets, metrics, and baselines.
SnakeODiagnose Angular runtime errors using DevTools, ng.probe, and console utilities.
nikhilsitaramBenchmark Claude Code skill revisions with isolated runs and statistical grading.
charly-vibesExtract reusable prompt templates from AI conversations by analyzing successful patterns.
