Security audit
Security checks for vulnerabilities and agentic risk
This video-watermarking skill needs review because its optional replacement mode can delete originals or overwrite another video despite its safety promises.
Keep backups, preview the batch, and use a separate output folder outside the source directory. Avoid --replace until destination collision checks and failure-safe replacement are implemented. The evidence supports a data-loss concern, not malicious intent.
Without declared permissions the skill's intent is opaque and cannot be validated.
subprocess module calls execute external commands. Without careful input validation, this enables command injection.
def run(cmd):
return subprocess.run(cmd, capture_output=True, text=True,
encoding="utf-8", errors="replace",
creationflags=CREATION_FLAGS)
Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.
def run(cmd):
return subprocess.run(cmd, capture_output=True, text=True,
encoding="utf-8", errors="replace",
creationflags=CREATION_FLAGS)
When --replace changes the container extension, the code deletes the source before committing the validated output with os.replace. If that commit fails or execution is interrupted, the original is already gone; the replacement can also overwrite an existing same-stem destination file without checking, making batch processing capable of unintended data loss.
No suspicious patterns detected.