respond-to-issue

Warn

Audited by Snyk on May 3, 2026

Risk Level: MEDIUM
Full Analysis

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 reads GitHub issue title/body and comments using gh issue view ... on the NVIDIA/Megatron-LM repo, which ingests untrusted, user-generated content that the agent interprets to decide responses and next actions, creating a clear vector for indirect prompt injection.

Issues (1)

W011
MEDIUM

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

Audit Metadata
Risk Level
MEDIUM
Analyzed
May 3, 2026, 07:06 AM
Issues
1
Security Audit — snyk — respond-to-issue