Scribe

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

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.

Type: OpenClaw Skill Name: scribe Version: 1.0.0 The skill is classified as suspicious due to its extremely broad access to highly sensitive local user data, including all Cursor IDE chat history (from `~/Library/Application Support/Cursor/User/globalStorage/state.vscdb` and `workspaceStorage/*/state.vscdb`) and a wide range of draft files (e.g., `**/*draft*.txt`, `blog/**/*.md`) within the user's workspace, as implemented in `scripts/scribe.py`. While this extensive data collection is explicitly stated as the skill's purpose in `SKILL.md` and `README.md`, and there is no evidence of data exfiltration or malicious intent in the current code, the sheer volume and sensitivity of the data it processes locally represent a significant privacy risk and a large attack surface. Any future vulnerability in the skill or the OpenClaw environment could lead to severe consequences given the aggregated sensitive information.