Source Reliability & Contradiction Resolution

1. Source Reliability Tiers

Reliability Tier	Numerical Weight	Typical Publishers	Operational Role
VERY HIGH	0.95 – 1.00	CISA KEV, National CERTs, Primary Software Vendors, Cryptographic Commit Logs	Authoritative ground truth. Confirms factual existence and active exploitation.
HIGH	0.85 – 0.94	Recognized Security Research Labs (e.g. Google TAG, Citizen Lab, Pillar Security, Trend Micro)	Deep technical analysis, zero-day discovery, and binary reversing.
MEDIUM	0.70 – 0.84	Established Cybersecurity Journalism, Secondary Threat Intel Aggregators	Secondary corroboration and broader contextualization.
LOW	0.40 – 0.69	Unattributed Technical Blogs, Social Media Postings (X/Mastodon)	Triage leads; cannot trigger score escalations without primary corroboration.
VERY LOW	0.00 – 0.39	Anonymous Claims, Unverified Forums	Disregarded or quarantined under adversarial deception filters.

2. Contradiction Resolution Protocol

When multiple credible sources publish conflicting assessments:

Rule 1 (Never Silent Compromise): Hermes never secretly averages scores or silently accepts the latest claim.
Rule 2 (Explicit Contestation): The relationship state is set to CONTESTED.
Rule 3 (Divergence Audit): An alert box is published displaying the conflicting claims side-by-side with source attributions.

┌────────────────────────────────────────────────────────┐
│ ⚠️ INTELLIGENCE DISAGREEMENT: CONTESTED STATUS         │
├────────────────────────────────────────────────────────┤
│ Source A (Security Researcher): Zero-day weaponization │
│ observed in the wild.                                  │
│ Source B (Primary Vendor): Telemetry indicates only    │
│ non-functional scanning attempts.                      │
├────────────────────────────────────────────────────────┤
│ Hermes Assessment: Active weaponization is plausible   │
│ but unconfirmed. Threat score held at 78 pending KEV.  │
└────────────────────────────────────────────────────────┘