<p>Configure automatic self-improvement triggers for AI coding agents.</p>

<h3>Overview</h3>

<p>Hooks enable proactive learning capture by injecting reminders at key moments:</p>
<ul>
<li><strong>UserPromptSubmit</strong>: Reminder after each prompt to evaluate learnings</li>
<li><strong>PostToolUse (Bash)</strong>: Error detection when commands fail</li>
</ul>

<h3>Claude Code Setup</h3>

<h4>Option 1: Project-Level Configuration</h4>

<p>Create <code>.claude/settings.json</code> in your project root:</p>

<pre><code>
{
"hooks": {
"UserPromptSubmit": [
{
"matcher": "",
"hooks": [
{
"type": "command",
"command": "./skills/self-improvement/scripts/activator.sh"
}
]
}
],
"PostToolUse": [
{
"matcher": "Bash",
"hooks": [
{
"type": "command",
"command": "./skills/self-improvement/scripts/error-detector.sh"
}
]
}
]
}
}
</code></pre>

<h4>Option 2: User-Level Configuration</h4>

<p>Add to <code>~/.claude/settings.json</code> for global activation:</p>

<pre><code>
{
"hooks": {
"UserPromptSubmit": [
{
"matcher": "",
"hooks": [
{
"type": "command",
"command": "~/.claude/skills/self-improvement/scripts/activator.sh"
}
]
}
]
}
}
</code></pre>

<h4>Minimal Setup (Activator Only)</h4>

<p>For lower overhead, use only the UserPromptSubmit hook:</p>

<pre><code>
{
"hooks": {
"UserPromptSubmit": [
{
"matcher": "",
"hooks": [
{
"type": "command",
"command": "./skills/self-improvement/scripts/activator.sh"
}
]
}
]
}
}
</code></pre>

<h3>Codex CLI Setup</h3>

<p>Codex uses the same hook system as Claude Code. Create <code>.codex/settings.json</code>:</p>

<pre><code>
{
"hooks": {
"UserPromptSubmit": [
{
"matcher": "",
"hooks": [
{
"type": "command",
"command": "./skills/self-improvement/scripts/activator.sh"
}
]
}
]
}
}
</code></pre>

<h3>GitHub Copilot Setup</h3>

<p>Copilot doesn't support hooks directly. Instead, add guidance to <code>.github/copilot-instructions.md</code>:</p>

<pre><code>

Self-Improvement

After completing tasks that involved:

• Debugging non-obvious issues
• Discovering workarounds
• Learning project-specific patterns
• Resolving unexpected errors

Consider logging the learning to .learnings/ using the format from the self-improvement skill.

For high-value learnings that would benefit other sessions, consider skill extraction.
</code></pre>

<h3>Verification</h3>

<h4>Test Activator Hook</h4>

<ol>
<li>Enable the hook configuration</li>
<li>Start a new Claude Code session</li>
<li>Send any prompt</li>
<li>Verify you see <code><self-improvement-reminder></code> in the context</li>
</ol>

<h4>Test Error Detector Hook</h4>

<ol>
<li>Enable PostToolUse hook for Bash</li>
<li>Run a command that fails: <code>ls /nonexistent/path</code></li>
<li>Verify you see <code><error-detected></code> reminder</li>
</ol>

<h4>Dry Run Extract Script</h4>

<pre><code>
./skills/self-improvement/scripts/extract-skill.sh test-skill --dry-run
</code></pre>

<p>Expected output shows the skill scaffold that would be created.</p>

<h3>Troubleshooting</h3>

<h4>Hook Not Triggering</h4>

<ol>
<li><strong>Check script permissions</strong>: <code>chmod +x scripts/*.sh</code></li>
<li><strong>Verify path</strong>: Use absolute paths or paths relative to project root</li>
<li><strong>Check settings location</strong>: Project vs user-level settings</li>
<li><strong>Restart session</strong>: Hooks are loaded at session start</li>
</ol>

<h4>Permission Denied</h4>

<pre><code>
chmod +x ./skills/self-improvement/scripts/activator.sh
chmod +x ./skills/self-improvement/scripts/error-detector.s

还有什么补充参数吗？一起完善。
</code></pre>