nvidia-skill-finder

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

• Third-party content exposure detected (medium risk: 0.65). SKILL.md instructs the runtime catalog check to fetch live catalog content via npx skills add nvidia/skills --list and fallback GitHub/build/raw URLs, whose skill names/descriptions are then matched and used to recommend prompts—those fetched pages/JSON are outsider-authored free text entering the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly instructs the agent at runtime to consult live catalog URLs (e.g. https://raw.githubusercontent.com/NVIDIA/skills/main/skills.sh.json) as the source of truth for recommending and naming skills, so fetching that remote catalog can directly control prompts/recommendations.