Ping Tool
Analysis
This is a simple ping utility with no evidence of malicious behavior, though it can send network probes and its metadata/documentation are slightly incomplete.
Findings (4)
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.
Checks for instructions or behavior that redirect the agent, misuse tools, execute unexpected code, cascade across systems, exploit user trust, or continue outside the intended task.
ping-tool [options] <hostname>
The skill lets the agent direct ping traffic to an arbitrary hostname. This is purpose-aligned for network diagnostics, but users should only test hosts they are authorized to contact.
subprocess.run(["ping", "-c", "4", sys.argv[1]])
The script depends on an external system ping binary, while the metadata declares no required binaries. This is expected for a ping tool but is an under-declared dependency.
subprocess.run(["ping", "-c", "4", sys.argv[1]])
The script launches a local system command. It does not use shell=True and the command is directly related to the stated ping purpose, but local command execution is still behavior users should notice.
- `-c N`: Send N packets then stop - `-i N`: Wait N seconds between packets - `-s N`: Set packet size - `-t N`: Set timeout in seconds
The documentation advertises several options, but the bundled script hard-codes four packets and only uses the first argument as the host. This is a documentation/capability mismatch rather than evidence of deception.
