blog-discourse
Warn
Audited by Snyk on Jul 24, 2026
Risk Level: MEDIUM
Full Analysis
MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).
- Third-party content exposure detected (high risk: 0.75). High likelihood: in Phase 3 the skill captures WebSearch SERP “snippet” text from outsider-authored public platforms (e.g., Reddit/HN/X/Medium) and then Phase 4/
render_markdownplaces those untrusted snippet strings into the LLM-visibleDISCOURSE.mdcontent that can later be read by other agents, creating an indirect prompt-injection path (snippets → results.json →scripts/discourse_research.py→ markdown output consumed by downstream LLM context).
Issues (1)
W011
MEDIUMThird-party content exposure detected (indirect prompt injection risk).
Audit Metadata