Reorder the install provider picker and add Grok Build, Amp Code, and omp

[luongnv89]

Type

improvement (medium confidence)

Description

The interactive install flow's provider picker (Step 2/8, "Selecting provider") presents providers in an order that does not match how users actually pick them, and it does not offer three providers the reporter wants available.

Current state:
The picker opens with Claude Code first, followed by Codex, OpenCode, Pi, Hermes, OpenClaw, Agents, Cursor, GitHub Copilot, Windsurf, Google Antigravity, Gemini CLI, Cline, Roo Code, and 5 more entries below the fold. Grok Build, Amp Code, and omp are not offered at all. The Agents entry carries no note about which harnesses support it.

Proposed change:
Reorder the list so it begins with:

1. Agents — annotated to say most harnesses support this one, except Claude Code
2. Claude Code
3. Pi
4. OpenCode
5. Codex

The remaining providers keep their current relative order after those five. Additionally, add three new providers to the list: Grok Build, Amp Code, and omp.

Related issues:
None identified.

Reporter Context
update the order of provider when installing skill to have the following order:

• Agents (and note that most of harnesses support this one, except Claude Code)
• Claude Code
• Pi
• Opencode
• Codex
  and the rest.
  Add also into list: grok build, amp code, omp

here is how it is right now: [screenshot below]

Screenshots

Acceptance Criteria

• The provider picker lists Agents, Claude Code, Pi, OpenCode, and Codex as its first five entries, in that order (high confidence)
• Every provider currently offered still appears after those five, in their existing relative order, with none dropped (high confidence)
• The Agents entry shows a note stating that most harnesses support it, except Claude Code (high confidence)
• Grok Build, Amp Code, and omp appear in the picker as selectable providers (high confidence)
• Select All / Deselect All and per-provider selection behave as they do today against the new order and the added entries (medium confidence)

Metadata

Priority: P2 (medium confidence)
Effort: M (medium confidence)
Suggested model: GPT-5.6 Sol High · Opus 5 Medium (CursorBench 3.2, 2026-09-02 — advisory)
Labels: enhancement (medium confidence)

[luongnv89]

Mid-flight requirement added by the reporter during resolution: on first-time setup (no saved preferences.selectedTools), the picker should pre-check Claude Code and Agents, not Agents alone. Implemented on the resolution branch.