intel: appsec research + community updates 2026-03-22

[kamalsrini]

Automated Skill Updates

Source: researchsecurityplan.md (2026-03-22) + communityfeedbackplan.md (2026-03-21)

Skills Updated

• secure-code-review (v1.0.0 -> v1.0.1)
• dependency-scanning (v1.0.0 -> v1.0.1)

Research Findings Applied

• ArXiv 2603.19138: LLM binary analysis patterns (knowledge-guided prioritization, early pruning)
• ArXiv 2603.18740: LLM confirmation bias in code review — exploitation via supply-chain attacks
• ArXiv 2603.18693: Cross-ecosystem vulnerability analysis (vendored libs, OS backport FP/FN)

Community Feedback Applied

• Research Signal 5 (HIGH): AI-generated PR vulnerability patterns (87% vuln rate)
• Community Improvement 2 (MEDIUM): IDE shift-left dependency scanning (GlassWorm campaign)

Human Review Required

Do not merge without review. Check line counts stay under 500.