@leonardocouy
I set up an online entry point for your writing-clearly-and-concisely skill. Anyone who lands on the repo can upload a draft and get Strunk-edited output without cloning or configuring anything.
The detail that stood out: you split the Elements of Style into loadable sections by concern -- grammar in 02, composition principles in 03, word choice in 05 -- so an agent can pull just the 1,000-4,500 tokens it needs instead of the full reference. That progressive disclosure is an opinionated design choice. Most writing-aid skills dump the entire style guide into context and hope the model sorts it out.
The AI-pattern removal list is the other half that makes the skill practical. Strunk tells you what to do; the Wikipedia-sourced signs-of-ai-writing.md tells you what to stop doing. Together they give the model two concrete checklists instead of a vague "write better" instruction.
Visitors can try it on their own draft here -- paste or upload a README, commit message, or report and see what the skill actually changes. Every run leaves a usage record, so you can see what kinds of text people bring and where the rules bite hardest. That signal is useful for deciding whether to expand the word-choice section or add more composition examples.
shesonglin@tinkerland.ai if you want the usage-record review link.
Feel free to close if this isn't relevant.
@leonardocouy
I set up an online entry point for your writing-clearly-and-concisely skill. Anyone who lands on the repo can upload a draft and get Strunk-edited output without cloning or configuring anything.
The detail that stood out: you split the Elements of Style into loadable sections by concern -- grammar in
02, composition principles in03, word choice in05-- so an agent can pull just the 1,000-4,500 tokens it needs instead of the full reference. That progressive disclosure is an opinionated design choice. Most writing-aid skills dump the entire style guide into context and hope the model sorts it out.The AI-pattern removal list is the other half that makes the skill practical. Strunk tells you what to do; the Wikipedia-sourced
signs-of-ai-writing.mdtells you what to stop doing. Together they give the model two concrete checklists instead of a vague "write better" instruction.Visitors can try it on their own draft here -- paste or upload a README, commit message, or report and see what the skill actually changes. Every run leaves a usage record, so you can see what kinds of text people bring and where the rules bite hardest. That signal is useful for deciding whether to expand the word-choice section or add more composition examples.
shesonglin@tinkerland.ai if you want the usage-record review link.
Feel free to close if this isn't relevant.