Knowledge Management Skill (Local Storage)

Organize your OpenClaw memory files into a structured local knowledge base. Automatically parses MEMORY.md and daily memory files, classifies entries by content type, and stores each as a timestamped markdown file in the appropriate folder.

Setup

No API keys needed! Just ensure your workspace structure exists:

mkdir -p ~/.openclaw/workspace/{Research,Decision,Insight,Lesson,Pattern,Project,Reference,Tutorial}

The skill will create folders automatically if missing.

Tools

km sync

Extract and store knowledge entries as markdown files.

Options:

• --days_back <num> (default: 7) — Include N days of daily files
• --dry_run — Preview without creating files
• --limit <num> — Max entries to process
• --cleanup — Delete orphan files (files in folders but not tracked in state)
• --verbose — Detailed logs

Example:

km sync --days_back 14 --dry_run
km sync --days_back 7 --cleanup

km classify

Parse and classify entries without storing. Outputs JSON with full metadata.

Options:

• --days_back <num> (default: 0) — 0 = only MEMORY.md, otherwise include daily files

Example:

km classify --days_back 3 > entries.json

km summarize

Create index files listing all entries in each content type folder.

Options:

• --output_dir <path> — Where to write index files (default: workspace root)

Example:

km summarize --output_dir ~/.openclaw/workspace

km cleanup

Remove orphaned files (files in content folders that are not tracked in the sync state).

Options:

• --dry_run — Preview what would be deleted

Example:

km cleanup --dry_run
km cleanup

km list_types

List all available content types and their descriptions.

Example:

km list_types

Storage Structure

~/.openclaw/workspace/
├── Research/
│   ├── 20260215T1448_OpenClaw_Optimization_Findings_e4b30e75.md
│   └── 20260215T1448_Benchmark_Results_9de25742.md
├── Decision/
│   └── 20260215T1448_Multi_Agent_Architecture_Choice_8fb1b6ab.md
├── Insight/
│   └── 20260215T1448_Cost_Savings_Realization_3c76995.md
├── Lesson/
│   └── 20260215T1448_Security_Misconfiguration_Fix_401aa1e.md
├── Pattern/
│   └── 20260215T1448_Workflow_Automation_Trend_e73f2a4.md
├── Project/
│   └── 20260215T1448_NFT_Liquidity_Skill_Plan_9349f18.md
├── Reference/
│   └── 20260215T1448_OpenClaw_API_Cheatsheet_a70f85d.md
├── Tutorial/
│   └── 20260215T1448_Setup_Automated_Backups_Guide_7a41f8e.md
├── MEMORY.md
├── memory/
│   ├── 2026-02-15.md
│   ├── 2026-02-14.md
│   └── local-sync-state.json   (tracks content_hash → filepath)
└── Research_Index.md
    Decision_Index.md
    Insight_Index.md
    ... (other index files)

Workflow Diagram

flowchart LR
    A[MEMORY.md<br/>Daily memory/*.md] --> B[EntryClassifier<br/>Parse & Classify]
    B --> C{Check State<br/>local-sync-state.json}
    C -->|Already synced| D[Skip]
    C -->|New entry| E[Write .md file<br/>timestamp_title_hash.md]
    E --> F[Update State<br/>content_hash → filepath]
    F --> G[Index Generation<br/>*_Index.md]
    G --> H[Organized Knowledge Base]

Loading

The diagram shows the complete sync pipeline from source memory files to the final organized folder structure with indexes.

File Naming

Each entry gets a timestamp prefix with a content hash suffix:

YYYYMMDDTHHMM_Title_With_Underscores_HASH.md

Example: 202602151440_GitHub_Documentation_Standardization_Protocol_e4b30e75.md

• Timestamp: 20260215T1440 (compact, sortable)
• Title is sanitized (invalid filesystem chars → _, spaces → _)
• 8-character content hash suffix prevents collisions when titles are identical
• Ensures chronological sorting while preserving distinct entries

File Content

Each .md file uses YAML frontmatter for metadata:

---
title: "GitHub Documentation Standardization Protocol"
content_type: "Research"
domain: "OpenClaw"
certainty: "Verified"
impact: "Medium"
confidence_score: 8
tags: ["AI", "Benchmark", "Automation"]
source: "MEMORY.md"
source_file: "MEMORY.md"
date: "2026-02-11"
content_hash: "e4b30e75d0f5a662"
---

Body content starts here...

How It Works

1. Parse MEMORY.md and recent daily memory/*.md files
2. Classify each entry: content type, domain, certainty, impact, confidence score, tags
3. Deduplicate using content hash (MD5 of title+date+body_snippet)
4. Check state — skip already-synced entries
5. Write file to appropriate folder with timestamp + hash filename
6. Update state — record content_hash → filepath mapping
7. Cleanup (optional) — delete files no longer in sync state

Classification Logic

• Content Type: Determined by keywords in title/body (Research, Lesson, Decision, Pattern, Tutorial, Reference, Insight)
• Domain: AI Models, OpenClaw, Cost Optimization, Trading, Learning, Process, General
• Certainty: Verified, Likely, Speculative, Opinion (based on phrasing)
• Impact: High, Medium, Low, Negligible (based on importance indicators)
• Tags: Auto-extracted from keyword map (AI, Automation, Cost, OpenRouter, etc.)
• Confidence Score: 1-10 based on source, length, data mentions

Customize these in the EntryClassifier class in index-local.js.

State Management

The sync state (memory/local-sync-state.json) tracks which entries have been exported:

{
  "e4b30e75d0f5a662": "/home/ubuntu/.openclaw/workspace/Research/202602151440_GitHub_Documentation_Standardization_Protocol_e4b30e75.md",
  "a70f85d63fa7f842": "/home/ubuntu/.openclaw/workspace/Reference/202602151440_GitHub_Collaborator_Access_Policy_a70f85d.md"
}

This enables:

• Idempotent syncs (no duplicates)
• Fast check before writing
• Easy cleanup of orphaned files

Do not edit this file manually unless you know what you're doing.

Troubleshooting

"Unknown tool" or "km: command not found"

• Ensure ~/workspace/bin is in your PATH, or call via full path: /home/ubuntu/.openclaw/workspace/bin/km
• Run npm link in the skill directory to register globally

No entries found

• Check your MEMORY.md structure: must have ## section headers, then ### entry titles under recognized sections
• Recognized sections: Research, Lesson, Decision, Pattern, Tutorial, Reference, Insight, etc.

Files not created

• Check write permissions on workspace folders
• Run with --verbose to see detailed logs
• Use --dry_run to see what would happen without writing

Old entries not syncing

• They may already be in sync state. Clear memory/local-sync-state.json to force full re-sync (will recreate all files with new timestamps)
• Better: Use --dry_run first to see skip messages

Duplicate files with same name

• This was a known issue (pre-Feb 15 14:50). Re-sync with --cleanup to regenerate with hash suffixes.

Cron Integration

Automate regular syncs to keep your local knowledge base current.

Via OpenClaw cron:

openclaw cron add \
  --name "Local Knowledge Sync" \
  --expr "0 */6 * * *" \
  --sessionTarget isolated \
  --payload '{"kind":"agentTurn","message":"km sync --days_back 7 --cleanup"}'

Via system crontab:

0 */6 * * * export WORKSPACE=/home/ubuntu/.openclaw/workspace && /home/ubuntu/.openclaw/workspace/bin/km sync --days_back 7 --cleanup

---

Version: 2.0.0 Changed: 2026-02-15 — Switched from Notion to local file storage, added hash suffixes for filename uniqueness Author: Claire (OpenClaw Agent) License: MIT