DevAgentRadar → Cline

You are helping me adopt a real coding-assistant change. Work only from the facts below. Do not invent features.

Context

Assistant: Cline Proposal: Pin Cline models after cli-v3.0.62 DeepSeek V4.1 Flash default swaps Summary: Cline CLI v3.0.62 ships Cline Desktop, Hub-managed Agent Plugins under ~/.agents/plugins, and a catalog refresh that changes unpinned defaults for 44 providers (many to DeepSeek V4.1 Flash). Pin models, keep plugins out of the repo, and install Desktop from cline.bot/desktop. Primary source: https://github.com/cline/cline/releases/tag/cli-v3.0.62

Why it matters

cli-v3.0.62 changes what runs if you leave Cline on implicit defaults. The bundled catalog grew from 5,788 to 6,079 models and the resolved default changed for 44 providers; most land on DeepSeek V4.1 Flash, while Gemini/Vertex resolve to Gemini 3.8 Flash, GitHub Copilot/Vivgrid to GPT-6 Astra, and NVIDIA to GLM 5.3 Flash. Unpinned sessions will silently switch models, context math, and cost. Agent Plugins are now Hub-managed: only ~/.agents/plugins/* is discovered and validated, skills are exposed as plugin-name:skill-name, and their MCP servers start without cline_mcp_settings.json. Workspace .agents/plugins is deliberately not scanned, so a cloned repo cannot start MCP servers. Cline Desktop is the new native app for open-weight models, Claude Code/Codex task import, scheduled runs, web search, voice input, and the marketplace; the CLI shows a one-time cline.bot/desktop notice unless CLINE_DISABLE_CLINE_PASS_NOTICE=1. Cline Pass now defaults to a subscribed model, shows as a configured provider after sign-in, and lists live-catalog models. Transient mid-stream provider errors retry up to 3 times with backoff (never after partial stream). Pin models and relocate plugins before those defaults take effect.

Suggested actions

1. Upgrade to Cline CLI v3.0.62, then open https://cline.bot/desktop and install Cline Desktop so you can import Claude Code and Codex tasks, run schedules, use web search and voice input, and browse the marketplace for plugins, MCP servers, and skills.
2. Install Agent Plugins only under ~/.agents/plugins/<plugin-name> (never in the repo's .agents/plugins). Confirm Hub lists them separately from Cline Plugins, their MCP servers start without editing cline_mcp_settings.json, and skills are invoked as plugin-name:skill-name.
3. Pin an explicit model on every Cline provider you use. Unpinned defaults now change for 44 providers: Hugging Face, Fireworks, Requesty, Nebius, Cortecs, CrossModel, DigitalOcean, Eden AI, and OpenCode Go resolve to DeepSeek V4.1 Flash; Gemini and Vertex to Gemini 3.8 Flash; GitHub Copilot and Vivgrid to GPT-6 Astra; NVIDIA to GLM 5.3 Flash.
4. If you do not want the CLI's one-time Cline Desktop startup notice, add export CLINE_DISABLE_CLINE_PASS_NOTICE=1 to your shell profile (at most one notice appears per launch otherwise).
5. After Cline Pass sign-in, open the model picker, confirm Pass appears as a configured provider, and select a subscribed-tier model if you are a subscriber (Pass now defaults to subscribed instead of whichever free model shipped last).

Config surfaces this release may change

• MCP servers — check your repo before applying
• skills — check your repo before applying

After you finish

Do not report this as applied to DevAgentRadar. You cannot write the visitor's loop.

Tell the human: open https://devagentradar.com/proposals/cline-cli-v3-0-62-pin-cline-models-after-cli-v3-0-62-deepseek-v4-1-flash and mark Applied, Skipped, or Failed. Proposal id: 5e8188c0-8f75-4725-a450-092f05cf71d5

Your job

1. Restate the change in one sentence.
2. Propose a minimal plan for my repo (or a throwaway pilot).
3. Implement only what I approve; prefer small diffs and tests.
4. Call out risks (permissions, breaking APIs, cost).

Start by confirming you understood the proposal.