Static findings for weak cryptography are false positives; the cited lines contain description text and argparse setup, not cryptographic algorithms. The skill still has confirmed high-risk behavior because it encourages black-box script execution and includes a helper that runs user-supplied server commands through a shell.
with_server.py passes the --server argument to subprocess.Popen with shell=True. A crafted server string can execute arbitrary shell operations in the caller environment.
The code explicitly uses shell=True with an argument supplied through the command line. The behavior is intentional, but it is still a clear arbitrary command execution surface.
with_server.py runs the remaining command-line arguments after servers are ready. This is expected for the tool, but it can execute unsafe commands if prompt input is untrusted.
The command is executed without shell=True, which reduces injection risk. It still runs arbitrary user-selected programs as part of normal operation.
Example scripts write console logs and screenshots to local paths. This is useful for debugging but can store sensitive page content outside the project.
The writes are explicit and local, so this is not exfiltration. The risk is moderate because captured browser content can include private data.
SKILL.md instructs agents to avoid reading helper source before running it and to treat scripts as black boxes. This can suppress inspection of executable code before use.
The text directly tells the agent not to read source before executing helper scripts. It is suspicious in a marketplace skill because it can interfere with source inspection and safety review.
The detected URLs point to localhost and are used as example targets for local web testing. No external data exfiltration endpoint is present.
All cited URLs use localhost and match the stated purpose of testing local web applications. I did not find evidence of remote collection or transmission.
The cited lines contain descriptive metadata and argparse option definitions, not weak hashing or cryptographic operations. No evidence found for cryptographic misuse.
The cited lines were reviewed directly and contain no cryptographic API usage. This is a scanner pattern mismatch.