Proofreader

校对/错别字检查、语法纠错、文风统一、一致性检查、可读性评分、校对报告。Proofreading with typo detection, grammar correction, style consistency, readability scoring.

This skill appears to do exactly what it claims: generate proofreading prompts and provide local helpers. It does not request credentials or perform network calls. Important things to consider before installing: - The scripts create and write files under $PROOFREADER_DIR (or $XDG_DATA_HOME/proofreader or $HOME/.local/share/proofreader) including data.log and history.log. If you plan to proofread sensitive text, that text may be saved locally; set PROOFREADER_DIR to a secure location or remove those files after use. - The skill can be invoked autonomously by the agent (disable-model-invocation is false), which is normal for skills. If you are concerned about autonomous runs, disable autonomous invocation in your agent settings. - The code is simple shell scripts with no obfuscation and no external network activity. Continue to review updates or new versions, and avoid supplying secrets or private credentials to this skill (none are required).

Type: OpenClaw Skill Name: proofreader Version: 2.0.0 The proofreader skill bundle is a legitimate tool designed to assist with text校对 (proofreading), grammar checking, and style consistency. The core logic in `scripts/proofread.sh` generates structured prompts for an AI agent to perform linguistic analysis, while `scripts/script.sh` provides basic local logging and data management utilities. No evidence of data exfiltration, malicious command execution, or harmful prompt injection was found.