Menu
Use at the start of EVERY session. Loads the user's personality profile and operates according to it. Also enables continuous distillation of new signals.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | principle |
| description | Use at the start of EVERY session. Loads the user's personality profile and operates according to it. Also enables continuous distillation of new signals. |
Two functions: (1) operate according to the user's principles, (2) continuously distill new ones.
Read ~/.claude/principles.md at session start. From that point forward, follow the user's documented patterns as your default behavior:
The principles file IS your instruction manual for this user. Treat it with the same weight as a system prompt.
After each user message, silently assess: did this reveal something new? If yes, update ~/.claude/principles.md. Do not announce, do not ask.
Distill when they:
~1 in 5 messages has a signal. Most are task-oriented noise.
Capture this person specifically. The test: could another agent read the file and behave indistinguishably from one that has worked with this user for weeks?
Write as principles (abstract patterns), not case examples (specific incidents). But specific enough to predict behavior in new situations.
Write personal blog posts with the rhythm, structure, and voice of top-tier essayists (Paul Graham, Dan Luu, Henrik Karlsson, Ben Kuhn, etc.)
Add a new feature with design, TDD, and PR workflow
Audit a codebase for security vulnerabilities, code quality issues, and architecture concerns
Use when the user says "autopilot", "auto mode", "替我回答", or "自动回答". Enables autonomous decision-making by answering Claude's own questions based on the user's personality profile.
Fix a bug with systematic debugging, TDD, and PR workflow
Create a git release — tag, push, and create GitHub release
