design-auditor

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). This is a third‑party personal GitHub repository (user Ashutos1997) rather than an established/official project — there’s no direct executable link shown but installing via npx/skills or cloning the repo could execute unreviewed code, so it’s potentially suspicious and should be code‑reviewed before use.

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

• Third-party content exposure detected (high risk: 0.70). The required workflow is installing/invoking the upstream design-auditor skill from a public GitHub repo, and at runtime the agent will execute that outsider-authored skill code which may ingest free-form text from external sources (e.g., user-provided design content or fetched pages) into the LLM context; this SKILL.md stub itself doesn’t show any exclusion that guarantees no outsider text is ingested.

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

• Potentially malicious external URL detected (high risk: 0.90). The SKILL includes explicit install commands that use npx to fetch and run the package Ashutos1997/claude-design-auditor-skill (https://github.com/Ashutos1997/claude-design-auditor-skill), which downloads and executes remote code at install/runtime and can therefore directly control agent behavior.