v3.1.0

Added

• Added
  tune-mobile-client-performance
  for evidence-led tuning of a specific SwiftUI or React Native interaction
  across motion, observation, rendering, media, network, and observability
  boundaries, with six execution evals and twenty balanced routing cases.

Documentation

• Separated general iOS capability decisions from mobile performance tuning:
  advanced pinned standalone XcodeBuildMCP into an agent-tooling pilot, kept
  Memgraph and ETTrace task-gated, and selected a first-party App Intents skill
  pilot from NextUp's existing shipped implementation rather than installing or
  forking build-ios-apps.

Fixed

• Made the skills.sh freshness monitor honor bounded Retry-After guidance and
  retry transient HTTP failures so catalog throttling is not mistaken for
  stale skill content.

Documentation

• Recorded Patrick Skills v3.0.0 release and exact-tag installation evidence,
  downstream Patrick Plugins v3.4.0 and Agent Tooling v0.8.0 verification, the
  remaining skills.sh stale-snapshot gate, and the requirement to dogfood
  bootstrap-repository only on the next genuinely new or empty real repository.