User Preferences
Set defaults for tone, style, and output once. Reuse forever.
At Car-Mart Technology Group: preferences fall into three categories — code (language, framework, style, tests, forbidden patterns), content (voice, format, length, banned phrases, Car-Mart vocabulary), and communication (response length, ambiguity handling, confirmation behavior, pushback tone).
Concrete Car-Mart examples: Dev sets 'Python 3.12 + black + ruff, no raw SQL' in api-refactor Project. Data sets 'tables-first, SQL with comments, list assumptions' in q3-dashboard Project. Ops sets 'SRE tone, lead with severity + hypothesis + next action' in runbook-update Project.
The full long-form read with the 3-category breakdown, the where-they-live table, the starter template, the common pitfalls, and the 3-preference-minimum lives at hool.dev/aitraining/preferences (the companion doc, ~5-7 min).
I'm a [role] at Car-Mart Technology Group (seniority, primary language, and one Project in scope: api-refactor / q3-dashboard / runbook-update / other). Suggest 5 Claude Preferences I should set as my default — across the 3 categories (code, content, communication). For each: setting name, the exact value, and the reason. End with the 1 preference I should set first if I only have 30 seconds.In your lane
Dev lane: T3 sets Preferences for terse + code-first + error-handling rationale. T4 writes a CLAUDE.md for the api-refactor Project with the 3 most-referenced helper modules, the team's PR template, and 'flag any new dependency not in requirements.txt.' Concrete artifact: open Desktop → Settings → Preferences → set 'terse default' + 'show test cases' + 'always include file path' → open Project: api-refactor → new chat → 'review this PR diff per our PR template' → notice the response matches the template before you ask it to.
Data lane: T3 sets Preferences for tables-first + SQL with comments + list 3 edge cases. T4 builds a Project for q3-dashboard with the 2 data dictionaries and the 1 sample stakeholder summary as Project knowledge. Concrete artifact: open Desktop → Project: q3-dashboard → Preferences (inherited from global) → new chat → 'explain this query, then list the 3 edge cases it does not cover, then propose a test' → notice the table is first, prose second.
Ops/Security lane: T3 sets Preferences for SRE tone + lead with severity/hypothesis/next action + confirm before destructive actions. T4 builds a Project for runbook-update with the runbook template and a Project instruction that flags any escalation-path or on-call-rotation change. Concrete artifact: open Desktop → Project: runbook-update → new chat → drop in the new runbook → 'compare to template, flag deltas, suggest the next runbook to update' → notice the response leads with severity, not preamble.
Set 3 preferences now (terse default + 5 banned phrases + pick-and-flag ambiguity). Open one Project (api-refactor / q3-dashboard / runbook-update). Run a 1-sentence task. Notice the next response is closer to your voice than any chat you started cold. Then save your preferences file to: (a) your team's Slack #ai-training, (b) your personal prompt library, (c) the 90-day plan export. The directors and CTO Josh are in the room to see what your team can ship today.