Repository navigation
Replies: 1 comment
|
Note Grok responding on behalf of Julius. Thanks for the detailed write-up and the screenshots. This is the same ask as #11307, which wants applied skills visible without expanding the collapsed tool group. I'm closing this one as a duplicate so the conversation stays in one place. Please add your always-visible row design and your local implementation notes there, since a working approach for Claude would help that discussion move forward. Claude skill calls already show the skill name since #14980, but they are still grouped. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Problem
When Claude loads a skill, the
Skillcall ends up inside the collapsed "Used N tools" group. To see which skill loaded, you expand the group, then expand the "Skill: fleet" row inside it, and all you get is raw JSON. It's barely visible like dimmed as well.[[
]]
A skill changes the instructions the agent works from, and Claude can load one without being asked. When the agent does something unexpected, the loaded skill is the first thing I check. It is a must to see engine loads specific skill, that is why in the stock Claude Code CLI shows it on every call:
[[
]]
Proposal
Give each Claude
Skillcall its own row that is always visible, the way the CLI does:[[
]]
Skill(name)with its arguments, and "Successfully loaded skill" under itI have this working in a local build. It touches 3 files in
apps/web(+190/-1) and follows the existing context compaction row inderiveMessagesTimelineRows. Typecheck and format are clean. I have not checked dark theme, narrow windows, or providers other than Claude.Questions
If the direction works for you, I will open a PR linked to this discussion. The code was written with Claude Opus 5.5 on xhigh in T3 Code.
All reactions