Skip to content

[FEATURE]: Collapse skill body in user message #33980

Closed

Description

@Arcadi4

Feature hasn't been suggested before.

  • I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

Currently, OpenCode appends the full skill/command body to the user message. This takes up a large space in the UI and makes it difficult to navigate in the session history.

It would be great if we could have a dedicated UI component for skills/commands used. Would be better if they could be toggled to collapse/show just like generic tool outputs now.

Example from the Codex app:

Image

Activity

  1. changed the title [-][FEATURE]: Folding command or skill body in user message[/-] [+][FEATURE]: Hiding skill body in user message[/+] on Jun 26, 2026
  2. iWorld-y commented on Jul 27, 2026

    @iWorld-y

    I’m running into the same issue.

    The Codex (currently ChatGPT) app provides a much cleaner experience here: invoking a skill only adds a lightweight skill reference or metadata entry to the visible conversation, while the agent can still access the full skill instructions internally.

    In OpenCode, I currently avoid invoking skills explicitly. Instead, I mention the skill name in a normal message and let the agent load it automatically. This keeps the conversation readable, but it is an inconvenient workaround.

    Ideally, explicit skill invocation should:

    • display only the skill name and description in the user message;
    • keep the complete SKILL.md content available to the agent;
    • optionally allow users to expand the full content;
    • avoid rendering the skill instructions as if they were written by the user.

    In other words, the visible conversation representation should be separated from the context supplied to the agent.

  3. changed the title [-][FEATURE]: Hiding skill body in user message[/-] [+][FEATURE]: Collapse skill body in user message[/+] on Jul 27, 2026
  4. Arcadi4 commented on Jul 27, 2026

    @Arcadi4
    ContributorAuthor

    I’m running into the same issue.

    The Codex (currently ChatGPT) app provides a much cleaner experience here: invoking a skill only adds a lightweight skill reference or metadata entry to the visible conversation, while the agent can still access the full skill instructions internally.

    In OpenCode, I currently avoid invoking skills explicitly. Instead, I mention the skill name in a normal message and let the agent load it automatically. This keeps the conversation readable, but it is an inconvenient workaround.

    Ideally, explicit skill invocation should:

    • display only the skill name and description in the user message;
    • keep the complete SKILL.md content available to the agent;
    • optionally allow users to expand the full content;
    • avoid rendering the skill instructions as if they were written by the user.

    In other words, the visible conversation representation should be separated from the context supplied to the agent.

    Unfortunately, they don't usually merge UI/UX PRs, so the best thing to do would be to mention a maintainer, @rekram1-node.

  5. mibrahimdev commented on Aug 21, 2026

    @mibrahimdev

    Any news?

  6. Arcadi4 commented on Sep 27, 2026

    @Arcadi4
    ContributorAuthor

    Shipped in v2!

  7. mibrahimdev commented on Sep 27, 2026

    @mibrahimdev

    Finally, thank you

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions