not-matthias/extract-knowledge
modules/home/programs/cli-agents/shared/skills/extract-knowledge/SKILL.md
Extract patterns and challenges from past Claude Code conversation logs and convert them into reusable skills or memory entries. Run periodically to capture recurring solutions.
$ install --global
skillsauthnpx skillsauth add not-matthias/dotfiles-nix extract-knowledgeInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 20, 2026, 6:03 AM186.5s1 file scanned
SKILL.md
- name:
- extract-knowledge
- description:
- Extract patterns and challenges from past Claude Code conversation logs and convert them into reusable skills or memory entries. Run periodically to capture recurring solutions.
- license:
- MIT
- context:
- fork
- agent:
- Explore
Knowledge Extraction from Claude Code Logs
Extract recurring challenges, solutions, and patterns from past Claude Code sessions and distill them into skills or memory entries.
When to Use This Skill
- After intensive development sessions on a new project
- When you notice solving the same problem repeatedly
- Weekly scan to capture recent learnings
- Before a major refactor (capture current best practices first)
Log Location & Format
~/.claude/projects/<path-encoded-dir>/
├── sessions-index.json # Metadata index for all sessions in this project
└── <uuid>.jsonl # Individual session (JSON Lines)
sessions-index.json — use this to quickly find relevant sessions:
[{
"sessionId": "uuid",
"firstPrompt": "what the user asked",
"summary": "auto-generated summary",
"messageCount": 42,
"created": "2025-01-01T00:00:00Z"
}]
Extraction Workflow
1. Find relevant sessions
# List all projects
ls ~/.claude/projects/
# Scan session index for a project
cat ~/.claude/projects/<project>/sessions-index.json | \
jq '.[] | {prompt: .firstPrompt, count: .messageCount}' | head -50
# Filter by topic
cat ~/.claude/projects/<project>/sessions-index.json | \
jq '.[] | select(.firstPrompt | test("error|fix|debug|fail"; "i")) | .firstPrompt'
2. Extract user messages from a session
cat ~/.claude/projects/<project>/<session>.jsonl | \
jq 'select(.type == "user") | .message.content' 2>/dev/null | head -30
3. Categorize findings
| What you found | Target skill |
|----------------|--------------|
| Nix hash errors, build failures, overlays | nix-package |
| IDA SDK patterns, plugin boilerplate | ida-plugin-dev |
| CodSpeed integration, benchmark setup | codspeed-bench |
| Binary analysis, unpacking, emulation | reverse-engineer |
| Rust cargo patterns | rust |
4. Write the finding
Use this format when adding to a skill:
### [Problem Title]
**Context**: When does this happen?
**Symptom**: What error/behavior do you see?
**Fix**:
```code```
**Why**: Brief explanation
5. Where to put it
- Skill file (
modules/home/programs/cli-agents/shared/skills/<name>/SKILL.md): reusable techniques, non-obvious gotchas, command patterns - Memory (
~/.claude/projects/<dir>/memory/): project-specific context, preferences
What NOT to Extract
- One-off hacks for a specific version
- Things already in CLAUDE.md
- Long context dumps — summarize to the essential pattern only
Related Skills
not-matthias/pi-agent
tools
VerifiedTrustedCommunity
Spawn the pi coding agent with a specific model/provider. Use when asked to run pi with a particular model, switch pi's model, use DeepSeek V4 Flash/Pro in pi, or look up pi's --model/--provider/--models CLI flags and thinking-level shorthand.
9SKILL.mdUpdated Jun 3, 2026
not-matthias/zoxide-nav
development
VerifiedTrustedCommunity
Navigate to directories using zoxide (frecency-based directory jumper). Use when the user says "go to", "navigate to", "cd to", "jump to" a project or directory by nickname/partial name (e.g. "go to my dotfiles", "jump to dot").
9SKILL.mdUpdated Apr 20, 2026
not-matthias/zellij
tools
VerifiedTrustedCommunity
Use when manipulating Zellij sessions, creating tabs or panes, sending commands to panes, capturing output, or looking up Zellij CLI commands for terminal multiplexer operations
9SKILL.mdUpdated Apr 20, 2026
not-matthias/writing-style
development
VerifiedTrustedCommunity
Emulates not-matthias's technical blog writing style. Use when writing blog posts, technical articles, README content, or any long-form technical prose. Produces investigation-driven, first-person narratives with dry humor, practical code examples, and concrete takeaways.
9SKILL.mdUpdated Apr 20, 2026