[CLI MCP] Make `ig ai-config` agent-agnostic instead of hard-coding `.claude/skills`

[IMinchev64]

Framework

All / Not framework-specific

Related CLI Command

ai-config

Feature Area

CLI command improvement

Problem Statement

ig ai-config currently treats .claude/skills as the default and effectively canonical destination for AI skill files. That is too tool-specific. .claude is valid for Claude Code, but the Ignite UI CLI is positioning this as general AI tooling, and projects may use GitHub Copilot, Codex, Cursor, Windsurf, or other assistants with different conventions and setup requirements.

Because of that, the current behavior creates a mismatch between the command’s intent and its implementation:

• the implementation is Claude-specific
• the user-facing concept is broader AI tooling
• some bundled docs already acknowledge other agent locations, but the actual copy logic still writes only to .claude/skills

This should be addressed as a generalization task, not as a Claude-only workflow improvement.

Proposed Solution

Refactor the ai-config flow so skill installation is agent-aware rather than hard-coded to .claude/skills.

The task should cover:

• introducing an explicit concept of AI target/client into ig ai-config
• supporting multiple destinations depending on the selected tool
• keeping .claude/skills as one supported option, but not the only one
• allowing either:
  • a selected target such as claude, copilot, codex, cursor, windsurf, generic
  • or a custom destination path for tools without a fixed convention
• maybe possibly aligning the generated docs/help text with the actual behavior?

The implementation should also review whether the MCP/config side of ai-config is similarly biased toward a single environment and make that explicit or configurable where needed.

Alternatives Considered

No response

Additional Context

The task should account for:

• Claude Code
• GitHub Copilot
• Codex
• Cursor
• Windsurf
• generic/custom setups

[Marina-L-Stoyanova]

Hi, thank you for the feedback, we have an open feature request for this : 1652 , but I will keep yours too as it is very descriptive.