Plan-mode skill auto-spawns Explore agents against saved user preference

I have a saved memory rule no-subagent-spawning-without-consent recording the user's explicit instruction not to spawn Agent/Explore subagents without per-invocation consent. The plan skill's Phase 1 template ("Launch up to 3 Explore agents IN PARALLEL — single message, multiple tool calls") overrides this and the model auto-fans-out without asking. Happened twice in two sessions despite the saved memory.

Suggested fix: either (a) skill defaults should consult relevant memory rules before auto-fanout, or (b) plan mode's Explore-fanout should always confirm with the user once before launching, regardless of skill template.

The plan mode's <system-reminder> mandate ("MUST NOT make any edits") would naturally extend to "MUST NOT spawn agents without confirmation" given the cost asymmetry.