web-publish

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly requires reading and passing credentials (e.g., -Password and -UserName) from .v8-project.json and examples show embedding a plaintext password into generated PowerShell commands, forcing the LLM to handle and output secret values verbatim.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). The skill's scripts explicitly download and extract a public Apache distribution from https://www.apachelounge.com/... (see scripts/web-publish.ps1 and scripts/web-publish.py), ingest and read files like httpd.conf from that archive, and then run the downloaded httpd.exe—so it fetches and executes untrusted third-party content at runtime which can materially change behavior.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill's runtime downloads and installs a portable Apache binary from https://www.apachelounge.com/download/VS18/binaries/httpd-2.4.66-260131-Win64-VS18.zip (and may execute its httpd.exe), which is a runtime-fetch of remote executable code required for the skill to operate.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill instructs the agent to modify Apache configuration (httpd.conf), start the Apache service and potentially download/install Apache — actions that change system files and services and typically require elevated privileges, so it can compromise the machine state.