Add skill invocation via slash commands #1623
Conversation
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
There was a problem hiding this comment.
Review Summary
Found 2 confirmed bugs in the skill invocation feature:
- HIGH severity: Skill file content is corrupted when displayed to the agent
- MEDIUM severity: Skill loading errors are silently swallowed without user feedback
Both issues are in the new ResolveSkillCommand and ResolveInput functions in pkg/app/app.go.
Skills can now be invoked as /commands in the TUI, both through the command palette and by typing /skill-name [args] directly. When skills are enabled for the current agent, they appear in a "Skills" category in the command palette. Skill resolution follows the same pattern as agent commands: the App resolves /skill-name to a prompt string by reading the SKILL.md file, then the chat page sends it as a regular message to the agent. Assisted-By: cagent
Signed-off-by: David Gageot <david.gageot@docker.com>
| git commit -m "bump <module_path> from <old_version> to <new_version>" -m "" -m "Assisted-By: cagent" | ||
| ``` | ||
| Record the dependency as **bumped** in your tracking table. | ||
|
|
||
| - **If either fails**: revert all changes and move on: | ||
| ```sh | ||
| git checkout -- . |
There was a problem hiding this comment.
I'm generally strongly against agents committing for me or generally manipulating git too much, but lets see if others feel the same.
If users don't use a good enough model, disasters are bound to happen 😅
There was a problem hiding this comment.
since this is not reusable anywhere outside of this repo, should this even be a skill or rather something our agent just knows how to do? 🤔
we could probably compact the prompt quite a bit, include it in the system prompt and get the same results.
There was a problem hiding this comment.
yes, we could but I want to play a bit more with it.
- I think on Claude we can call tools from skills to make them resolve in a single round-trip. I'd like to play with that
- We don't all use the same agent but we could be using the same set of skills
Also, by activating this as a skill, I realised the things we were missing (/command support and schema)
2 participants
No description provided.