[Feature Request] support agent skills

[lightaime]

Motivation

We will supprt agent skills in https://github.com/camel-ai/camel ChatAgent and then support agent skills in eigent based on ChatAgent`

Solution

No response

Alternatives

No response

Additional context

No response

[jcmrs]

Just please be careful not to fall prey to the same trappings Gemini CLI fell into.

Skills are methodologies. Not scripts or executables.

https://github.com/google-gemini/gemini-cli/issues/15895

[Pakchoioioi]

Phototype Demo

https://github.com/user-attachments/assets/f6d313ab-c66f-4d00-85bf-47a7302daa93

Skills feature _ more detailed explanation

Allow users to add new Skills by uploading a skill package (.zip) and configure its activation scope (Global or specific Agent/Worker). Support quick testing and deletion management.

0) Entry & Navigation (Homepage > Capabilities > Skills)

Provide a fixed in-product entry point for Skills by adding a Capabilities tab to the top main navigation of the Homepage (parallel to Projects / MCP & Tools / Browser / Settings).

Navigation Path: Homepage → Capabilities → Skills

After clicking Capabilities, users enter the Capabilities page framework:

Left-side secondary navigation (within Capabilities) includes:

• Models (could be moved up to the previous level.)
• Skills (new/activated entry in this requirement)
• Memory

Default behavior:

• When users enter Capabilities, they land on Models by default. if we put "Models" on the homepage top bar, then "skills" by default.

Skills Page Display (Right Main Content Area)

When users select Skills within Capabilities, the right main content area displays the Skills management page, including:

• Page title Skills and description text (Repeatable, customizable instructions… Learn more)
• Search box
• + Add upload entry
• Tabs: Your skills / Example skills (current version supports Your skills only)

Skill list items:

• If not empty, display name, description, added time, enable toggle, scope display, overflow (···) menu, etc.
• If empty, display “No skills added by you yet”

1) Upload Entry & Interaction

On the Skills page, click + Add to open the Upload skill modal.

The modal supports drag-and-drop upload and file selection.

File Requirements

• The uploaded .zip (or skill package) must include a SKILL.md file in the root directory.
• SKILL.md must define the following information using YAML:
  • name (skill name)
  • description (skill description)

Text: File Requirements The uploaded .zip or skill package must include a SKILL.md file located in the root directory. The SKILL.md file must define a skill name and description using YAML format.

After a successful upload:

• The skill appears in the Your skills list.
• Displays name, description, added time, and other basic information.
• The skill is enabled by default.

2) Skill Scope (Activation Range)

Each skill provides a Skill Scope dropdown configuration. Default value: Global Selectable options include (Multi-select):

• Global
• Developer Agent
• Browser Agent
• Multi-modal Agent
• Document Agent
• (All Workers configured by the user, Newly created Workers must automatically appear in the selectable list.)

Scope Rules: Global: Effective for all Agents/Workers. Partially Available (Multi-select) Effective only for selected Agents/Workers. Let me clarify more at here: in the future, when we support automatic skill generation, the skills generated by a specific agent will automatically default to the scope of that particular agent. For example, skills saved after the Browser Agent completes a task will automatically be assigned to the Browser Agent. Users will be able to modify this in the Skills settings page.

3) Availability & Auto-Disable Rules

Each skill provides an Enable/Disable toggle on the right side. When the target bound to the Skill Scope changes:

If a bound Worker is deleted:

• If the skill is still bound to other targets → Remains enabled.
• If no targets remain bound after deletion → Automatically set the skill to disabled (toggle off).

4) Skill Action Menu (···)

Click the ··· menu on the right side of a skill to expand options.

Try in chat

Action:

• Creates a new Project.
• Prefills a guided prompt, for example:

I just added the /skills_name skill for Eigent, can you make something amazing with this skill?

Delete (Destructive Action – Red)

Action:

• Opens a secondary confirmation window.
• After confirmation, deletes the skill (it can be re-added later by re-uploading).

5) Delete Confirmation Modal

Title: Delete this skill? Description: Indicates that the skill can be re-added by uploading again.

Buttons:

• Cancel
• Delete (Confirm deletion, red)

Acceptance Criteria (Summary)

• Users can upload .zip files and validate that SKILL.md exists and contains parsable YAML (name/description).
• Newly added skills default to Scope = Global and are enabled by default.
• Scope supports Global and multi-target selection; newly created Workers automatically appear as selectable options.
• When a Worker is deleted:
  • If no bound targets remain, the skill is automatically disabled.
  • Otherwise, it retains its original state.
• The ··· menu includes Try in chat and Delete, with Delete requiring secondary confirmation.
• Try in chat successfully assigns the task to an enabled agent and executes it.