yifeizhen

Daily Reflection — AI skill for Claude Code

AI community

Daily self-examination (每日省察) for AI-native work: an agent skill that scans ALL your coding-agent sessions (Claude Code / ZCode / Codex) each night, judges them against your goal registry, and emails.

How to install Daily Reflection

This entry records only its repository, not the path inside it, so there is no exact command to give. Open yifeizhen/daily-reflection and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Daily Reflection does

Alternatives in AI

EchoBird — One-click install + model switch:Claude Code,Codex CLI (OpenAI), Grok Build (xAI), DeepSeek Harness, Kimi Code 3.1k ★
Orkas — Open-source multi-agent AI desktop client — build and command your AI agent team through conversation 1.5k ★
Agor — Orchestrate Claude Code, Codex, and Gemini sessions on a multiplayer canvas 1.2k ★

README

Daily Reflection · 每日省察

**One agent skill. Every night it reads ALL your AI coding sessions — Claude Code, ZCode, Codex — judges them against *your* goal registry, and emails you a reflection report: what you did, what's left, and exactly 3 concrete next steps.**

[![License: MIT](https://img.shields.io/badge/License-MIT-5b4fc9.svg)](LICENSE) [![Made with a smack](https://img.shields.io/badge/made%20with-a%20painless%20smack-c0392b)](#the-story)

📖 **[See a sample report →](examples/demo-report.html)**

The story

「日日省察，方日日见财。」— Examine yourself daily, and every day yields wealth.

In middle school I was a poor student with bad grades. My Chinese teacher **Chen Guoping** never wrote me off — 有教无类, every student counts. He preached daily self-examination, pushed us to read more, and delivered the most memorable *painless smack* I've ever seen. (He'd have made a great CI system: instant feedback, zero malice.)

In 2026 the work changed: I don't write code alone anymore — I run **dozens of agent sessions a day** across Claude Code, ZCode and Codex. Output is no longer the bottleneck; **direction is**. Sessions pile up; goals quietly drift.

So I turned teacher Chen's lesson into a skill: every night, one automated painless smack — every goal gets a verdict, every claim points back to a session file — and the smack always comes with candy: exactly 3 concrete suggestions.

What it does

Each run:

Frames your goals from goals.md — a tiny goal registry you own (deadlines, last-pushed dates, long-term anchors).
Scans every session store on the machine — Claude Code (~/.claude/projects), ZCode (~/.zcode/cli/db/db.sqlite), Codex (~/.codex/sessions) — extracting per session: project, intent, user turns, time span. Empty stores are recorded, never skipped.
Judges each goal: pushed ✅ / stalled ⚠️ / drifting 🔴 — with a facts-only rule: every "did today" line must trace back to a session file, or it doesn't get written.
**Emails y