fix(gmail-auth): request explicit filter settings scope

docs/plans/state.json

@@ -3,6 +3,20 @@
   "updated_at": "2026-02-23",
   "description": "Tracks the status of all Flynn plans and implementation phases",
   "plans": {
+    "gmail-filter-scope-correction": {
+      "status": "completed",
+      "date": "2026-02-23",
+      "updated": "2026-02-23",
+      "summary": "Corrected Gmail OAuth scopes for filter creation after validating that `mail.google.com` alone still produced `ACCESS_TOKEN_SCOPE_INSUFFICIENT` for `users.settings.filters.create` in live checks. `flynn gmail-auth` now explicitly requests `gmail.settings.basic` (plus `gmail.readonly`). Updated auth tests and operator docs/comments.",
+      "files_modified": [
+        "src/cli/gmail-auth.ts",
+        "src/cli/gmail-auth.test.ts",
+        "README.md",
+        "config/default.yaml",
+        "docs/plans/state.json"
+      ],
+      "test_status": "pnpm test:run src/cli/gmail-auth.test.ts + pnpm typecheck passing"
+    },
     "minimal-tui-multiline-paste-mode": {
       "status": "completed",
       "date": "2026-02-23",
@@ -6345,7 +6359,7 @@
     "operator_dx_milestone": "Phase 3 (Live Ops Dashboard): 2/2 plans complete — milestone done",
     "dashboard_observability": "completed — service health graphs + core service log viewer added to web UI via observability RPCs and bounded backend sampling",
     "gmail_auth_cli": "flynn gmail-auth command implemented with OAuth2 flow, doctor check, config routed to Telegram",
-    "gmail_filter_creation": "completed — gmail.filter.create tool added with criteria/action validation; gmail-auth now requests full Gmail scope (https://mail.google.com/) for complete filter permissions",
+    "gmail_filter_creation": "completed — gmail.filter.create tool added with criteria/action validation; gmail-auth requests explicit gmail.settings.basic + gmail.readonly scopes for filter creation and inbox reads",
     "toolloop_action_intent_recovery": "completed — when a model claims it will execute a tool but emits no tool call, NativeAgent now issues one internal nudge and continues the same turn to execute tools or produce a concrete blocker",
     "toolloop_execution_claim_recovery": "completed — when a model claims a known tool already succeeded/failed without emitting a tool call, NativeAgent now nudges once and retries the same turn before returning text",
     "minimal_tui_multiline_paste_mode": "completed — minimal TUI now supports `/paste`/`/multiline` multiline compose mode ending with single '.' line, preventing newline truncation for pasted prompts",