Word Reader
PassAudited by ClawScan on May 1, 2026.
Overview
This skill coherently reads user-selected Word documents, but users should review the install steps and avoid processing sensitive documents unless they are comfortable exposing that text to the agent.
This appears appropriate for reading Word files. Before installing, review the dependency installation commands, preferably use a virtual environment, and only run it on documents whose contents and metadata you are willing to share with the agent.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
If you process confidential Word files, their text, tables, and metadata may become visible to the agent conversation or saved output.
The skill is designed to extract document contents and metadata, including in batch mode, so private or untrusted document text may be returned into the agent context or output.
元数据获取 - 读取文档属性（作者、标题、创建时间等） ... python3 {baseDir}/scripts/read_word.py <目录路径> --batchUse it only on documents you intend the agent to read, and treat extracted document text as data rather than instructions.
Running the installer can alter your Python or system package setup.
The setup script installs external packages without version pinning and may use system package managers; this is expected for the parser but changes the local environment.
python3 -m pip install python-docx --break-system-packages ... sudo apt install -y python3-pip
Review the install script first and consider installing dependencies in a virtual environment.
Processing .doc files depends on a local external converter being installed and trusted.
For legacy .doc support, the skill invokes the local antiword converter on the user-selected file; no shell interpolation is shown.
subprocess.run(['antiword', str(self.file_path)], capture_output=True, text=True, encoding='utf-8')
Install antiword only from a trusted package source, and keep document paths user-selected.