Repository navigation
Conversation
PR #46 told both talk-like-scarletkc and x-content to add angle, attitude, jokes and hooks whenever drafting, and dropped the old guard against turning every output into an emotional opinion post. Thank-you and apology posts came out overworked. - Limit the "add angle, attitude, reaction" permission to opinion, review and complaint posts; thank-you, apology, status, announcement and personal-feeling posts are written plainly, with the closing thanks and one expressive emoji allowed. - Remove the "say it sharper than the source" escalation from the voice rules; stance is kept at the user's strength, not raised. - Split the tweet section into an opinion register and a plain register, and exempt plain posts from the hook checklist and the "would anyone stop scrolling" reread in x-content. - Narrow both descriptions so short thank-you, congratulation and greeting posts no longer auto-load the skills. - Add evals 44 and 45 (talk-like) and 29 (x-content) covering a thank-you polish, a delay apology and a thank-you draft.
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
The short thank-you exclusion in the description also caught requests that ask for the author's voice without naming the skill, such as the eval 44 prompt. Let an explicit voice request trigger it; the plain register in the body now handles those posts correctly.
The thank-you post that motivated the emoji exception was itself AI-generated; the author did not want the emoji. Restore the original "no decorative emoji" rule with no carve-out for thank-you posts. Eval 44 now asks the skill to bring an overdone AI thank-you draft back to plain speech, and evals 44, 45 and 29 are cut to three behavior-level assertions each.
There was a problem hiding this comment.
🔍 Emoji exception conflicts with stated scope
The PR description promises to preserve user-supplied emoji, but this rule removes that exception. The thank-you evaluation now requires deleting the supplied 🙏. Confirm which behavior the skill is intended to teach.
(Refers to this code)
Was this helpful? React with 👍 or 👎 to provide feedback.
Owner
Author
|
Superseded by #49, which rewrites talk-like-scarletkc as a short voice reference and removes x-content. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation
#46 told both
talk-like-scarletkcandx-contentto add angle, attitude, jokes and hooks whenever drafting ("把话说得更有劲", "起草时可以说得更鲜明", "情绪说满"), and dropped the old guard "不要把所有输出都变成情绪化推文". Every post was pushed toward an opinion post. A thank-you post to subscribers came out overworked, and the skill rules would have deleted its closing "真的谢谢你们 🙏" as a decorative emoji and a padded ending.Changes
talk-like-scarletkcinto an opinion register and a plain register.x-content, exempt plain posts from the six-hook checklist ("找不到就换素材"), from choosing a stance for the author, and from the "would anyone stop scrolling" reread.talk-like-scarletkc44 (thank-you polish keeps the closing and emoji), 45 (delay apology stays plain);x-content29 (thank-you draft does not hunt for a hook).Compatibility surface
Both skill
descriptionfields change, which changes auto-triggering: short thank-you, congratulation and greeting posts no longer load either skill unless named. Chat messages still triggertalk-like-scarletkcas before, since the Chat register is corpus-backed and unaffected by this bug. README summaries are generated frommetadata.summary, which is unchanged.Verification
Eval JSON validated for unique ids (talk-like 40 evals, x-content 27).
End-to-end check: a fresh agent read only the revised SKILL.md files and wrote the three new eval prompts, plus a control written under the old #46 wording. All three drafts satisfy the new assertions. The delay-apology pair shows the difference:
New rules (eval 45 prompt):
Control under the old "主动加角度、态度、反应、吐槽、类比和反问，把话说得更有劲。情绪说满" wording:
The thank-you polish (eval 44) kept the original "真的谢谢你们 🙏" and only touched two phrases; the x-content thank-you draft (eval 29) opened on the milestone and ended on the thanks without a hook.