skills

This causes Claude to guess what grilling means and it asks all questions at once and does not provide options or recommendations. I tried tweaking the skill to explicitly tell it to load both skills and then to invoke them. This time it said it would load both skills, but then only loaded . I asked it about the other one and it said this:

I’d like to propose a new skill inspired by , but focused on performance improvements instead of architecture. The idea would be a skill that scans a codebase for potential performance bottlenecks, presents the findings clearly, and proposes possible solutions for improving them. Why this would be useful A lot of developers ask agents broad performance questions like: “Why is this app slow?” “Find performance bottlenecks in this codebase.” “How can I make this faster?” “Audit this app before launch.” “Where are the biggest performance issues?” Without a dedicated skill, agents may jump too quickly into random optimizations, caching, rewrites, or generic advice. A focused performance skill could help the agent approach this more thoughtfully by identifying likely bottlenecks, explaining why they matter, and giving the user a clear set of options to explore. Possible areas it could look at The skill could potentially surface issues around things like: slow database queries N+1 query patterns inefficient loops or data structures unnecessary re-renders large client bundles expensive synchronous work repeated network requests poor caching boundaries slow startup or cold-start behavior background job bottlenecks expensive work happening in hot paths How this differs from seems more suited for investigating a specific known issue or regression. This proposed skill would be more proactive: “Scan this codebase and show me the most important performance opportunities.” So it would be closer in spirit to , but with performance as the lens. Expected outcome The skill could produce a clear performance review of the codebase, including: what the possible bottlenecks are where they appear in the codebase why they may matter how severe they seem what kind of improvement may be possible what solution direction could be considered The main value would be helping developers quickly understand where performance work is most worth focusing, instead of guessing or optimizing randomly. Suggested names My preference is because it mirrors and makes the relationship clear.

per 16:10 , for skill with name grilling with doc https://github.com/mattpocock/skills/blob/main/skills/engineering/grill-with-docs/SKILL.md when it calls in productivity/grilling/SKILL.md written Interview me relentlessly about every aspect of this plan, but plan is not excising yet ? then i see contradiction grill-with-docs used before plan is created , but in order to use grill-with-docs/ plan is needed for - productivity/grilling/SKILL.md