#7398 chore(skills): remove dated patterns and redundancy from skill prompts

Summary

• Audited every skill under .agents/skills (plus add-settings-page) for instructions written for older models, stale paths/commands, and content duplicated across skills or already covered by CLAUDE.md and .claude/rules
• Fixed stale references: safeCompare import path, helm poll-cron shape, bun test → vitest invocation, CONNECTOR_META_REGISTRY icon read, deleted v2 rollout gate, cursorScopeKey, file-parse path, chunkArray source, getHostedModels provider list, provider flag consumption matrix, Exa hosted-key reference, defineWorkspaceOperation example missing the required capability
• Resolved skill-vs-rule disagreements: mutation onSuccess/onSettled, returned hook callbacks, read-once URL params, settings-page chrome ownership, block outputs nested example, credential oauth-input inference, boolean sub-block type, shadow-vs-border and press-scale guidance in the vendored design skills
• Collapsed duplicate copies to one owner: resolved-secret rules now live only in add-tools; add-integration points at add-tools/add-block/add-trigger instead of carrying drifted partial copies; validate-model reuses add-model's consumption matrix; babysit defers to ship's "What to Omit"
• Removed history narratives, migration-relative phrasing, pinned counts, and emphasis boosters; kept every tool contract, exact command sequence, and reasoned constraint

Type of Change

• Chore

Testing

bun run skills:sync --check, bun run lint, check-block-registry, and bun run check:audits (45 audits) all pass.

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)