PPT Translator
PassAudited by VirusTotal on May 11, 2026.
Overview
Type: OpenClaw Skill Name: ppt-translator Version: 1.0.0 The skill bundle's core script, `scripts/translate.py`, uses `subprocess.run` to execute the `libreoffice` command for rendering PowerPoint files. While this is a legitimate function for the stated purpose of the skill, it introduces a potential shell injection or path traversal vulnerability. If the OpenClaw agent were to pass unsanitized user-controlled input (e.g., file paths for `--input`, `--output`, or paths used by `libreoffice`) to this script, it could lead to arbitrary command execution or file system manipulation. There is no evidence of intentional malicious behavior such as data exfiltration, persistence, or prompt injection against the agent in `SKILL.md`.
Findings (0)
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.
Installation may require extra local packages that are not fully represented in the registry requirement fields.
The skill's own metadata requires local dependencies even though the registry summary says there are no required binaries and no install spec. Users need to install these dependencies from trusted sources.
"requires": {
"bins": ["libreoffice"],
"python": ["python-pptx"]
}Install LibreOffice and python-pptx from trusted package sources, and verify dependency versions before processing important files.
The skill will open/render presentations locally and create output files, which can consume local resources and write PNG/PPTX files to the chosen output directory.
The script launches LibreOffice to render the user-provided PPTX into PNG. This local command execution is central to the claimed layout-verification workflow and is not hidden.
subprocess.run(
['libreoffice', '--headless', '--convert-to', 'png',
'--outdir', output_dir, pptx_path],Use it only on PPTX files you intend to process, choose output paths deliberately, and keep LibreOffice updated.
Confidential slide text or rendered slide images could become part of the agent/model processing context.
The documented workflow may send extracted slide text and rendered slide images into LLM/Vision model processing. This is expected for translation and verification, but provider/data-boundary details are not specified.
Agent 提取所有文字 → 翻译（自己翻译或调用 LLM） ... Vision 模型检测文字溢出
Avoid using the skill on confidential presentations unless the configured model provider and data-retention policy are acceptable.