Problem
Feedback I've received from Venkat Subramaniam:
One nitpick I do have is the use of multiline lambdas http://blog.agiledeveloper.com/2015/06/lambdas-are-glue-code.html. It is easy to get drawn into that, but a little guidance to extract methods will be helpful for AI to not create multiline lambdas.
Area
Skill behavior
Describe the solution you want
Do an eval run of all the evals, analyze all the code results with skill, and check for usage of multi-line lambdas. Where they are used, add eval criteria to verify the LLM is not using them. Then adapt the skill until it achieves 100% with context on all evals, keeps the skill review quality at 100% and criteria is weighted appropriately. If any eval meets the defined bar, promote them. The goal is that the AI no longer uses multi-line lambdas anymore. Also consider adding more evals, potentially ones specific to multi-line lambdas if appropriate.
Examples
Alternatives considered
No response
Current workaround
No response
Additional context
Blog post in markdown format for easier processing: lambdas-are-glue-code.md
AI Assistance (if used)
AI prompts / session logs (optional)
Problem
Feedback I've received from Venkat Subramaniam:
Area
Skill behavior
Describe the solution you want
Do an eval run of all the evals, analyze all the code results with skill, and check for usage of multi-line lambdas. Where they are used, add eval criteria to verify the LLM is not using them. Then adapt the skill until it achieves 100% with context on all evals, keeps the skill review quality at 100% and criteria is weighted appropriately. If any eval meets the defined bar, promote them. The goal is that the AI no longer uses multi-line lambdas anymore. Also consider adding more evals, potentially ones specific to multi-line lambdas if appropriate.
Examples
Alternatives considered
No response
Current workaround
No response
Additional context
Blog post in markdown format for easier processing: lambdas-are-glue-code.md
AI Assistance (if used)
AI prompts / session logs (optional)