agent-browser
Browser automation using Vercel's agent-browser CLI. Use when you need to interact with web pages, fill forms, take screenshots,…
Show recent system improvements — captured learnings plus new Claude capabilities. Use when the user says 'what's new', 'any updates to how Dex works'. Not for previewing and applying a version update; use `dex-update`. Not for unused existing features; use `dex-level-up`.
Claude Code
$ npx -y skills add davekilleen/Dex --skill dex-whats-new --agent claude-code
Codex
$ npx skills add davekilleen/Dex --skill dex-whats-new -a codex -y
Cursor
$ npx skills add davekilleen/Dex --skill dex-whats-new -a cursor -y
Gemini CLI
$ npx skills add davekilleen/Dex --skill dex-whats-new -a gemini-cli -y
GitHub Copilot
$ npx skills add davekilleen/Dex --skill dex-whats-new -a github-copilot -y
OpenCode
$ npx skills add davekilleen/Dex --skill dex-whats-new -a opencode -y
Windsurf
$ npx skills add davekilleen/Dex --skill dex-whats-new -a windsurf -y
name: dex-whats-new description: "Show recent system improvements — captured learnings plus new Claude capabilities. Use when the user says 'what's new', 'any updates to how Dex works'. Not for previewing and applying a version update; use `dex-update`. Not for unused existing features; use `dex-level-up`."
Check for improvements to your Dex system from TWO sources: 1. **Learnings from your usage** - Mistakes, patterns, opportunities captured during sessions 2. **New Claude Code capabilities** - Features that could enhance your workflows
/dex-whats-new # Review learnings + check for Claude updates /dex-whats-new --full # Include detailed explanations /dex-whats-new --learnings # Only review session learnings /dex-whats-new --claude # Only check Claude Code updates
$MODE: Optional flags
---
**Check for learnings to review:**
1. Read files in `System/Session_Learnings/` from last 30 days 2. Read `06-Resources/Learnings/Mistake_Patterns.md` - check for new patterns 3. Read `06-Resources/Learnings/Working_Preferences.md` - check for trends
**Extract improvement opportunities:**
**Present findings:**
🧠 LEARNINGS FROM YOUR USAGE Since last review: [X] session learnings captured ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 📊 PATTERNS IDENTIFIED 1. [Pattern name] Observed: [frequency] times Impact: [what this affects] Suggestion: [concrete improvement] 2. [Pattern name] ... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 💡 SUGGESTED IMPROVEMENTS Based on what you've learned, here's how to improve Dex: 1. [Improvement name] Problem: [What's not working smoothly] Solution: [Specific change to make] Files: [What to update] Effort: Low/Medium/High 2. [Improvement name] ... Want me to implement any of these? (Enter number)
**If no learnings:** Skip this section or show: "No session learnings captured yet. The system will learn as you use it."
---
Load `System/claude-code-state.json`:
{
"last_check": "2026-01-15",
"last_version_seen": "1.0.28",
"features_seen": ["hooks", "sub-agents", "skills", "commands", "mcp"]
}If file missing or `last_check` is null, treat as first run.
Use `WebSearch` to find the latest Claude Code changelog:
Focus on:
Identify what's new since `last_check`:
**For each new feature:** 1. What it does (plain English, 1-2 sentences) 2. Why it matters for PKM users 3. How you could use it in Dex (concrete example) 4. Effort to adopt (Low/Medium/High)
**If Claude Code updates found:**
🆕 CLAUDE CODE UPDATES Last checked: [date] (X days ago) Current version: [version] ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 🆕 NEW FEATURES 1. [Feature Name] What: [Plain English description] For you: [How this could improve Dex] Effort: Low 2. [Feature Name] What: [Description] For you: [Specific improvement idea] Effort: Medium ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 💡 SUGGESTED IMPROVEMENTS Based on what's new, here are concrete things you could add to Dex: 1. [Improvement name] Uses: [Which new feature] What it does: [Specific description] Pillar: [Which pillar it supports] Want me to implement any of these? (Enter number) Or run `/dex-improve` to workshop custom ideas.
**If no Claude updates:**
✅ Claude Code is up to date! Last checked: Today Current version: [version] No new features since your last check.
---
When running without flags, show BOTH sections:
🔄 DEX SYSTEM IMPROVEMENT REVIEW ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 🧠 FROM YOUR USAGE [Session learnings section] ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 🆕 FROM CLAUDE CODE [Claude updates section] ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 💡 COMBINED IMPROVEMENTS Pulling together learnings + new capabilities: 1. [Improvement combining both] Uses: [Your pattern] + [New Claude feature] Impact: [What this unlocks] Want me to implement any of these? (Enter number or 'all')
**CRITICAL:** After presenting findings, you MUST update `System/claude-code-state.json` using the Write tool.
**Steps:** 1. Read current state file to preserve existing `features_seen` array 2. Add any newly discovered features to the array (avoid duplicates) 3. Update `last_check` to today's date (YYYY-MM-DD format) 4. Update `last_version_seen` if a new version was found 5. Write the updated JSON back to `System/claude-code-state.json`
**Example update:**
{
"last_check": "2026-01-28",
"last_version_seen": "1.0.32",
"features_seen": [
"hooks",
"sub-agents",
"skills",
"commands",
"mcp",
"parallel-sub-agents",
"async-hooks"
]
}**Verification:** After writing, confirm to user: "State file updated - next check will compare against today's findings."
---
After presenting findings to the user, automatically create or enrich backlog ideas:
1. **Call `synthesize_changelog()`** from Improvements MCP
One personal Dex, built around your work. Keep your priorities, people, meeting notes and tasks in a folder you own. Work with Dex through your AI app to plan the day, prepare for a meeting or follow through on a promise.
Browser automation using Vercel's agent-browser CLI. Use when you need to interact with web pages, fill forms, take screenshots,…
Build applications where agents are first-class citizens. Use this skill when designing autonomous agents, creating MCP tools,…
This skill should be used when writing Ruby gems following Andrew Kane's proven patterns and philosophy. It applies when creating…
This skill should be used before implementing features, building components, or making changes. It guides exploring user intent,…
Capture solved problems as categorized documentation with YAML frontmatter for fast lookup
Expert guidance for creating, writing, and refining Claude Code Skills. Use when working with SKILL.md files, authoring new…
Shape output for a reader with ADHD: lead with the next action, number multi-step work, restate state across turns, suppress…
Set up the owner profile so a personal agent knows who it works for (first-run setup, \"get to know me\"), or distill colleague,…