Scribe

Scans OpenClaw logs, config files, chat history, cursor history, behavior, desires, tastes, and drafts to take comprehensive daily and weekly notes with summ...

Security Scan

The skill's behavior (scanning logs, config, chat DBs, drafts and writing notes) matches its description, but it reads many highly sensitive local files (OpenClaw config, Cursor chat DBs, memory and behavior files) and ships executable code with no provenance — review before installing or running.

Purpose & Capability

The name/description (Scribe — comprehensive note-taking) matches the actions the skill performs: scanning logs, openclaw.json, Cursor SQLite DBs, memory and draft files and producing Markdown/JSON notes. There are no unrelated environment variables or external binaries declared, so capability requests are broadly coherent with the stated purpose.

Instruction Scope

SKILL.md explicitly instructs the agent (and the user) to read many user-local files and directories (e.g., ~/.openclaw/logs/*.log, ~/.openclaw/openclaw.json, Cursor's state.vscdb in the user Library path, workspace memory and drafts). While this is consistent with producing comprehensive notes, those instructions allow aggregation of very sensitive data (configuration that may contain tokens, chat transcripts, draft content, memory/behavior files). The SKILL.md also suggests cron integration which would automate repeated access to those sources.

Install Mechanism

No install spec is provided (instruction-only skill with a bundled script). That is lower-risk than a remote download/install, but the package includes a local Python script (scripts/scribe.py) that will be executed — there is executable code on disk with no declared provenance (git clone URL in README is a generic placeholder).

Credentials

The skill requires no environment variables, which is good, but it reads configuration files (openclaw.json) and Cursor DBs that commonly contain credentials, API keys, or private conversation history. Requesting broad filesystem access to aggregate these items is proportionate to a 'comprehensive scribe' only if the user expects that degree of access; otherwise it is high-risk. The skill does not declare where (if anywhere) collected data might be sent, so local aggregation could become an exfiltration target if the code is modified.

Persistence & Privilege

The skill does not request always:true and does not declare modifying other skills or system-wide agent settings. It writes output into workspace/Notes/* by design. The README shows an example cron job, but scheduling is optional and not enforced by the skill metadata.

What to consider before installing

Before installing or running Scribe: (1) Inspect scripts/scribe.py fully for any network activity (HTTP requests, sockets, remote endpoints, encoded/obfuscated strings) or subprocess calls that could exfiltrate data. Grep for modules like requests, urllib, socket, subprocess, or any hard-coded URLs. (2) Review your ~/.openclaw/openclaw.json and Cursor DBs to confirm they don't contain secrets you don't want aggregated; if they do, remove or rotate them first. (3) Run the script in an isolated environment or with --openclaw-home pointing to a copy of your data to see what it reads and what it writes. (4) If you plan to automate via cron, prefer running against a sanitized copy or explicitly restrict the openclaw-home path. (5) If the repository provenance (author, homepage, repo URL) is not verifiable, treat the included code as untrusted. What would change this assessment: discovering any network exfiltration (hard-coded remote endpoints, encrypted outbound traffic, or subprocesses that call external services) or obfuscated/encoded payloads would raise the severity toward malicious; a trusted upstream repository and an explicit statement that config files do not contain secrets (or that secrets are filtered) would increase confidence toward benign.

Scribe | OpenClaw Skill

Comprehensive note-taking system that scans multiple OpenClaw data sources to generate daily and weekly notes with summaries.

Description

Scribe automatically scans and analyzes:

• OpenClaw logs - Errors, warnings, gateway events, subagent activity
• Config files - Model preferences, gateway settings, agent configurations
• Chat history - Cursor IDE chat messages from SQLite databases
• Cursor history - Workspace-specific conversation history
• Memory files - Daily notes and long-term memory (MEMORY.md)
• Behavior files - BEHAVIOR.md, DESIRES.md, TASTES.md, PREFERENCES.md
• Drafts - Blog posts, tweets, and other draft files

Generates structured markdown notes with summaries at the top of each file.

Installation

clawhub install scribe

Or clone into your skills directory:

git clone https://github.com/Org/scribe.git workspace/skills/scribe

Usage

Daily Notes

Generate a daily note covering the last 24 hours:

python3 workspace/skills/scribe/scripts/scribe.py --mode daily

Weekly Notes

Generate a weekly note covering the last 7 days:

python3 workspace/skills/scribe/scripts/scribe.py --mode weekly

Both Daily and Weekly

Generate both notes at once:

python3 workspace/skills/scribe/scripts/scribe.py --mode both

JSON Output

Get results in JSON format:

python3 workspace/skills/scribe/scripts/scribe.py --mode daily --json

Examples

Example 1: Daily Note Generation
Scenario: You want a daily summary of OpenClaw activity.
Action: Run python3 workspace/skills/scribe/scripts/scribe.py --mode daily.
Outcome: A markdown file workspace/Notes/daily/YYYY-MM-DD.md with a summary at the top, followed by detailed sections for logs, chat history, memory, drafts, behavior, and config.

Example 2: Weekly Summary
Scenario: You want a weekly overview of patterns and trends.
Action: Run python3 workspace/skills/scribe/scripts/scribe.py --mode weekly.
Outcome: A markdown file workspace/Notes/weekly/YYYY-MM-DD.md with weekly statistics, trends, and activity summaries.

Example 3: Cron Job Integration
Scenario: Automate daily note generation.
Action: Add a cron job that runs scribe.py --mode daily every day at midnight.
Outcome: Daily notes are automatically generated and saved to workspace/Notes/daily/.

Commands

python3 workspace/skills/scribe/scripts/scribe.py --mode daily    # Generate daily note
python3 workspace/skills/scribe/scripts/scribe.py --mode weekly   # Generate weekly note
python3 workspace/skills/scribe/scripts/scribe.py --mode both     # Generate both
python3 workspace/skills/scribe/scripts/scribe.py --mode daily --json  # JSON output
python3 workspace/skills/scribe/scripts/scribe.py --openclaw-home /path/to/openclaw  # Custom home directory

• --mode — Choose daily, weekly, or both (default: daily)
• --json — Output results in JSON format instead of markdown files
• --openclaw-home — Specify OpenClaw home directory (default: ~/.openclaw)

What this skill does

1. Scans logs - Analyzes all .log files in logs/ directory for errors, warnings, gateway events, and subagent spawns
2. Reads config - Extracts model preferences, gateway settings, and agent configurations from openclaw.json
3. Extracts chat history - Queries Cursor's SQLite databases (state.vscdb) for recent chat messages
4. Scans memory files - Reads daily notes (memory/YYYY-MM-DD.md) and long-term memory (MEMORY.md)
5. Finds drafts - Searches for draft files matching patterns like *draft*.txt, *draft*.md, blog/**/*.md, tweet*.txt
6. Reads behavior files - Looks for BEHAVIOR.md, DESIRES.md, TASTES.md, PREFERENCES.md
7. Generates notes - Creates structured markdown files with summaries at the top
8. Saves to Notes - Writes daily notes to workspace/Notes/daily/YYYY-MM-DD.md and weekly notes to workspace/Notes/weekly/YYYY-MM-DD.md

Output Format

Daily Note Structure

# Daily Summary - YYYY-MM-DD HH:MM:SS

- **Logs**: X errors, Y warnings
- **Gateway Events**: Z events
- **Subagent Activity**: N spawns
- **Chat Messages**: M messages
- **Daily Notes**: K files
- **Drafts**: L draft files
- **Config**: Loaded successfully

---

# Daily Note

## Logs
### Errors (X)
- [error details]

### Warnings (Y)
- [warning details]

## Chat History
Found M messages in the last 24 hours.
### User (X messages)
- [message previews]

## Memory Files
### Daily Notes (K)
- **YYYY-MM-DD**: [content preview]

## Drafts
### [draft path]
- Size: X bytes
- Modified: [timestamp]
- Preview: [content preview]

## Behavior & Preferences
### Behavior Files
- **BEHAVIOR.md**: [content preview]

## Configuration
### Model Preferences
- Default: [model]
- Aliases: X configured

Weekly Note Structure

# Weekly Summary - YYYY-MM-DD HH:MM:SS

[Summary statistics]

---

# Weekly Note - YYYY-MM-DD to YYYY-MM-DD

## Weekly Summary
- **Total Errors**: X
- **Total Warnings**: Y
- **Gateway Events**: Z
- **Subagent Spawns**: N

## Chat Activity (M messages)
- **User**: X messages
- **Assistant**: Y messages

## Memory Activity (K daily notes)
- **YYYY-MM-DD**: X bytes

## Drafts (L files)
- **[path]**: X bytes (modified: [timestamp])

## Trends & Patterns
- ⚠️ **Error Rate**: X errors this week
- 🤖 **Subagent Activity**: N spawns this week
- 💬 **Chat Activity**: M messages this week

Requirements

• Python 3.7+
• OpenClaw installation with ~/.openclaw directory structure
• Cursor IDE installed (for chat history scanning)
• SQLite3 available (usually pre-installed on macOS)
• Write access to workspace/Notes/ directory

Integration as a Cron Job

Example Cron Job Configuration:

{
  "payload": {
    "kind": "agentTurn",
    "message": "Run scribe.py --mode daily to generate daily notes.",
    "model": "openrouter/google/gemini-2.5-flash",
    "thinking": "low",
    "timeoutSeconds": 300
  },
  "schedule": {
    "kind": "cron",
    "cron": "0 0 * * *"
  },
  "delivery": {
    "mode": "announce"
  },
  "sessionTarget": "isolated",
  "name": "Daily Scribe Note"
}

Or run directly via shell script:

# Add to crontab (crontab -e)
# Run daily at midnight
0 0 * * * /Users/ghost/.openclaw/workspace/skills/scribe/scripts/scribe.py --mode daily >> /Users/ghost/.openclaw/logs/scribe.log 2>&1

Security & Privacy

• File access: Reads openclaw.json (config only, no secrets), logs/*.log, memory/*.md, and Cursor SQLite databases
• No data exfiltration: All data stays local; notes are saved to workspace/Notes/ directory
• Safe execution: Uses read-only access to config and logs; only writes to workspace/Notes/daily/ and workspace/Notes/weekly/ directories

Limitations

• Chat history extraction depends on Cursor's database schema (may need updates if Cursor changes storage format)
• Draft file detection uses pattern matching; may miss files with non-standard naming
• Behavior file detection looks for common filenames; custom locations may not be found
• Large log files may take time to process