Manage persistent user preferences, corrections, and goals in a local JSON profile.
npx skills add https://github.com/qf-studio/navigator --skill nav-profile-qf-studio
Please help me install this Agent Skill. Skill: nav-profile Source: https://github.com/qf-studio/navigator/tree/main/skills/nav-profile Command: npx skills add https://github.com/qf-studio/navigator --skill nav-profile-qf-studio
.agent/.user-profile.json. - Auto-Learning from Corrections: Detects correction patterns like "No, I meant..." or "Not X, use Y" and silently saves generalized rules, keeping a rolling window of the last 20 corrections. - Goal Tracking: Records active work goals with status and last-mentioned dates, auto-archiving stale ones. - Use Case: After telling your AI assistant "Remember I prefer functional style and concise explanations", future sessions load those preferences automatically, and repeated corrections like "use plural REST endpoints" are learned without being asked twice. ## Quick Start Ask the assistant to save your preferences, for example by saying "remember I prefer concise explanations and functional code style".High-intent search queries and answers about installing and using this skill.
State your preference naturally, such as "remember I prefer functional style" or "I like concise explanations". The preference extractor maps common phrases to profile fields and writes them to .agent/.user-profile.json for future sessions.
The extractor watches for correction signals like "No, I meant...", "Not X, use Y", or "Always do X". When detected, it generalizes the correction into a pattern, stores it with a confidence level, and keeps only the most recent 20 corrections.
The profile lives at .agent/.user-profile.json, which is git-ignored and local only. It stores preferences, corrections, and goals but never code snippets, file contents, or conversation history, and you can delete it anytime.
The profile loader catches JSON decode errors and falls back to an empty profile with a warning instead of crashing. You can then recreate the profile with the create action or by saving new preferences.
Yes, the delete action removes the profile file entirely after confirmation. Future sessions then start fresh, and you can rebuild the profile by saving preferences as you work.
EhraxCreates isolated git worktrees for feature work with verified ignore rules and project setup.
wewetv1987-cellStore and reapply explicit coding preferences from user feedback.
Claude-Code-Community-IrelandOperate an MCP server for persistent, searchable AI agent memory.
caliber-ai-orgCapture user instructions as structured learnings and stage CALIBER_LEARNINGS.md.