Summary
This logs every failure, user correction, and capability gap to help Claude learn from mistakes across sessions. When a bash command times out, you say "actually, try it this way," or you ask for something it can't do, it writes structured logs that get reviewed at session start. The hook system catches corrections like "No, that's wrong" and integrates with QAVR to boost memory rankings for successful patterns. It's basically a feedback loop that persists beyond individual conversations. The capability tracker is especially useful because it shows you what features users keep requesting that don't exist yet. Fair warning, you'll need to maintain these log files or they'll grow indefinitely.
Install to Claude Code
npx -y skills add kimasplund/clawdbot-skills-pack --skill self-improving-agent --agent claude-codeInstalls into .claude/skills of the current project.
Keep your Mac awake while Claude Code and 40+ AI agents run. Sleeps when they're idle.
One time payment $9 →
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
On Capafy, your Skill runs online 24/7 as an agent product, and you get paid every time someone uses it.
Start earning →
Files
SKILL.mdView on GitHub
Self-Improving Agent
Captures learnings, errors, and corrections to enable continuous improvement across sessions.
Activation Triggers
This skill automatically activates when:
| Trigger | What Gets Logged | Location |
|---|---|---|
| Command fails | Error type, context, recovery suggestion | logs/failures_detailed.jsonl |
| User corrects | "No, that's wrong...", "Actually..." | logs/corrections.jsonl |
| Missing capability | "Can you X?" where X isn't available | logs/missing_capabilities.jsonl |
| API/tool fails | Failure pattern, suggested fix | logs/failures_detailed.jsonl |
| Better approach found | Optimization learned | logs/learnings.jsonl |
How It Works
1. Automatic Logging (via hooks)
PostToolUse → enhanced-failure-logger.js → logs failures with context
UserMessage → correction-detector.js → detects "wrong/actually/try again"
Response → capability-tracker.js → detects unfulfilled requests
2. Learning Aggregation
All learnings flow to logs/learnings.jsonl:
{
"timestamp": "2026-01-26T12:00:00Z",
"type": "user_correction|tool_failure|missing_capability",
"category": "factual_error|command_failed|web_browsing",
"description": "what was learned",
"source": "which detector"
}
3. Session Start Review
On each session start, recent learnings are shown:
=== Learning Review ===
[Learnings] 47 total entries
[Recent]
• [tool_failure] Bash failed: timeout - WebFetch to external API...
• [user_correction] User corrected: "No, use the other file..."
[Corrections] 12 user corrections logged
[Capability Gaps] Top requested:
• send emails (5x)
• browse web (3x)
4. QAVR Integration
Successful learnings boost Q-values for related memories, improving future retrieval.
Manual Commands
Review Learnings
# Show all learnings
cat ~/.claude/logs/learnings.jsonl | tail -20
# Show corrections only
cat ~/.claude/logs/corrections.jsonl | jq -s 'group_by(.correction_type) | map({type: .[0].correction_type, count: length})'
# Show capability gaps report
node ~/.claude/scripts/hooks/capability-tracker.js --report
Test Detection
# Test correction detector
node ~/.claude/scripts/hooks/correction-detector.js
# Test failure logger
node ~/.claude/scripts/hooks/enhanced-failure-logger.js
# Test capability tracker
node ~/.claude/scripts/hooks/capability-tracker.js
Learning Categories
Correction Types
factual_error- Wrong information providedretry_request- User asked to try againmisunderstanding- Misinterpreted the requestfailed_solution- Solution didn't work
Failure Types
permission_error- Access deniednot_found- File/resource missingtimeout- Operation timed outnetwork_error- Connection issuessyntax_error- Invalid syntaxapi_error- External API failedcommand_failed- Shell command failedagent_failed- Subagent failed
Capability Categories
web_browsing- Internet access requestsimage_processing- Image/photo handlingcommunication- Email/messagingdatabase_access- SQL/database queriesexternal_api- Third-party servicesmemory_persistence- Long-term memory
Configuration
In settings.json, these hooks enable self-improvement:
{
"hooks": {
"SessionStart": [...], // Reviews learnings
"PostToolUse": [
{"matcher": "Bash", "hooks": [{"command": "enhanced-failure-logger.js"}]},
{"matcher": "Task", "hooks": [{"command": "enhanced-failure-logger.js"}]}
]
}
}
Benefits
- Learn from mistakes - Don't repeat the same errors
- Understand user preferences - Track what corrections mean
- Identify skill gaps - Know what features to build
- Improve over time - QAVR ranking gets better with feedback
- Context persistence - Learnings survive session restarts
Integration with Other Skills
| Skill | Integration |
|---|---|
| QAVR | Successful learnings boost memory Q-values |
| Memory Consolidation | Periodic cleanup of old learnings |
| Confidence Check | Review learnings before major tasks |
| IR-v2 | Use learnings to inform pattern selection |
Files
~/.claude/
├── logs/
│ ├── learnings.jsonl # All learnings
│ ├── corrections.jsonl # User corrections
│ ├── failures_detailed.jsonl # Enhanced failure logs
│ ├── missing_capabilities.jsonl # Capability requests
│ └── capability_gaps.json # Aggregated gaps
└── scripts/hooks/
├── correction-detector.js
├── enhanced-failure-logger.js
├── capability-tracker.js
└── session-start.js (reviews learnings)
Featured
Keep your Mac awake while Claude Code and 40+ AI agents run. Sleeps when they're idle.
One time payment $9 →Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →On Capafy, your Skill runs online 24/7 as an agent product, and you get paid every time someone uses it.
Start earning →
