refresh-modelslisted

# Refresh Models The update leg of the choosing-models freshness loop. Detection compares data sources already in the session; **this command does the reasoning** — reading release notes and deciding what, if anything, the tier table should say differently. Never auto-apply threshold or doctrine edits; propose, then let the user approve. The data file lives in the sibling `choosing-models` skill of this plugin — resolve it as `${CLAUDE_PLUGIN_ROOT}/skills/choosing-models/models.toml` (fall back to the path relative to this file if the variable isn't set). Project overrides of that file are refreshed the same way, in their own location. ## Workflow 1. **Detect.** Run the check rather than performing it by hand: ```bash uv run --no-project -- python \ "${CLAUDE_PLUGIN_ROOT}/skills/choosing-models/scripts/lineup_check.py" <session model id> ``` It exits 1 and names the absent model when the session is running on something the tier data does not list — the environment tripwire, as a command. Then compare `models.toml` against the platform's model reference (e.g. the claude-api skill's current-models table, the session environment's lineup note where present, or the published models documentation via WebFetch) for the drift a single id cannot show: a current model missing from the table, a listed model no longer current, or `review_by` in the past. No drift and not past `review_by` → report "lineup current" and stop. 2. **Read the chang