[FEATURE]: Desktop Environment panel — skills, plugins, MCPs, loaded instructions + per-session context cost

[DankyoDev]

Problem

As an OpenCode Desktop user, I cannot see what is loaded into my sessions without reading opencode.json or asking the model (and hoping the answer is accurate). That breaks the Desktop promise: config-file spelunking for something the app already knows.

My own setup as a concrete example: 12 custom skills, 4 MCP servers (41 tools: codebase-memory 14, playwright 24, context7 2, grep.app 1), 4 plugins, ~12 KB of global instructions across 9 files. I only discovered half of this was actually loading by digging through opencode debug config and the server log — e.g. my skills lived in a directory OpenCode never reads, and nothing in the UI hinted at it.

What already exists / is in flight (not asking for that again)

• TUI /skills dialog; Web/Desktop parity tracked in [FEATURE]: Add /skills command and skill panel to Web/Desktop app #26297.
• Desktop MCP setup + connection testing: [FEATURE]: Add MCP server setup and connection testing to Desktop #40335 → PR feat(desktop): add MCP server setup and connection testing #43719.
• Plugins/LSP GUI + visible load failures: [Feature Request]: Desktop app needs GUI for managing plugins and LSP servers, plus visible load-failure errors #41037.
• TUI skills in status/sidebar: [FEATURE]: display loaded skills in status page and sidebar #7533 / feat: show skills in status and sidebar #6154.

What's missing: one read-only "Environment" view for Desktop

1. Skills — available (with source path) vs loaded in this session. Today: invisible unless you run opencode debug skill.
2. Plugins — loaded / failed + the error inline. Today: failures are log-only (background dependency install failed cost me a week of sluggish sessions before I found it in a 51 MB log).
3. Instructions actually injected — resolved list (global AGENTS.md + instructions[] + project overrides) with byte sizes. Today: no way to confirm which files won.
4. Per-session context cost breakdown — how much each of the above contributes per turn. Update: the bot pointed me to [FEATURE]: Session context usage (similar to /context in Claude) #6152, [FEATURE]: /context — detailed token/cost breakdown panel per session component #41908 and [FEATURE]:Show context usage in OpenCode Desktop, including for local OpenAI-compatible models #34900, which already cover this angle — conceding point 4 to those. This issue is now scoped to the inventory side (points 1–3), which none of them requests: what is loaded, from where, and whether it failed.

Proposing read-only v1 (a panel/tab next to the existing Server view). Editing/management can ride on #43719-style flows later. Happy to test early builds against a skill/MCP-heavy setup like mine.

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• [FEATURE]: Session context usage (similar to /context in Claude) #6152: Session context usage breakdown (similar to /context in Claude) — covers the per-session context cost breakdown aspect
• [FEATURE]: /context — detailed token/cost breakdown panel per session component #41908: /context detailed token/cost breakdown panel per session component — very similar context window visibility request
• [FEATURE]:Show context usage in OpenCode Desktop, including for local OpenAI-compatible models #34900: Show context usage in OpenCode Desktop, including breakdown by system prompt, tools, skills, and MCP metadata — closely overlaps the Desktop-specific panel you're describing

[DankyoDev]

Checked the three flagged issues (#6152, #41908, #34900): they cover the token/cost breakdown angle, so I'm conceding point 4 to them and narrowing this request to the environment inventory side (skills available-vs-loaded with paths, plugin load failures inline, resolved instructions list) — which none of those request.