oneshot

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

• Third-party content exposure detected (high risk: 0.75). Outsider free text can enter the LLM context when the user provides a GitHub issue reference/URL and the skill runs gh issue view ... --json title,body,labels, then uses the issue title/body (authored by an external party) as the task description for subsequent Explore/Code steps.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches GitHub issue content at runtime (e.g., via gh issue view or a GitHub issues URL like https://github.com/.../issues/N) and uses the issue title/body as the task description, which means external content directly controls the agent's instructions.