personalize-ai-assistantlisted

# Personalize AI Assistant ## Overview Use this skill to turn a user's repeated AI-assistant interactions into durable, auditable personalization: configuration rules, project/user instruction files, or a new custom skill. The goal is not to flatter the user; it is to make future assistants behave more consistently with verified preferences. ## Core Workflow 1. **Scope the target.** Identify whether the user wants global personalization, project-local behavior, a new skill, or a specific assistant config such as `AGENTS.md` or `CLAUDE.md`. If unspecified, default to a global user-level config plus a short evidence summary. 2. **Audit available surfaces.** Run `scripts/detect_personalization_context.py` when local files are relevant. Inspect existing config files before editing and preserve their intent. 3. **Collect behavior evidence.** Use recent thread summaries, user corrections, repeated requests, created artifacts, local project conventions, and existing skills/configs. Do not read sensitive histories such as shell history, mail, private chat exports, or browser data unless the user explicitly asks. 4. **Build an evidence ledger.** For each candidate preference, record the observed behavior, source, confidence, proposed rule, destination config, and risk. Read `references/personalization-model.md` for scoring and rule-writing patterns. 5. **Choose the packaging layer.** Read `references/config-targets.md` before editing assistant config files. Use a config rule for b