skills/shaun-z/auto-claude-code-research-in-sleep/monitor-experiment/Snyk

monitor-experiment

Warn

Audited by Snyk on Apr 10, 2026

Risk Level: MEDIUM

Full Analysis

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

Third-party content exposure detected (high risk: 0.70). The skill explicitly fetches and parses user-generated experiment data from Weights & Biases (Step 3.5: using wandb.Api() and referencing https://wandb.ai/... runs) and instructs the agent to interpret run summaries and histories to drive monitoring decisions, so untrusted third-party content could influence actions.

Issues (1)

W011

MEDIUM

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

Audit Metadata

Risk Level

MEDIUM

Analyzed

Apr 10, 2026, 02:14 AM

Issues

1