ghost-scan-deps

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches and executes external content and ingests public vulnerability data: agents/init/agent.md runs a curl from raw.githubusercontent.com to install wraith, agents/scan/agent.md runs the wraith scanner (which pulls OSV/GitHub data), and agents/analyze/analyzer.md and SKILL.md instruct the agent to read vulnerability summaries/references from the resulting candidates.json/OSV output and use them to drive analysis and remediation decisions, so untrusted third‑party content is read and can materially influence behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). The Init agent runs a runtime install command that fetches and executes remote code via curl -sfL https://raw.githubusercontent.com/ghostsecurity/wraith/main/scripts/install.sh | bash to install the required wraith binary, so this URL directly controls code executed by the skill.