Evidence that vetoes had no effects, and allows did

Evidence packs are delivered artifacts that demonstrate governance behavior at declared irreversible sinks: VETO ⇒ no side effects, ALLOW ⇒ side effects occurred—within the governed execution path.

📌 What an Evidence Pack is

• A structured bundle of artifacts tied to specific declared sinks.
• Designed for engineering review and internal assurance, not marketing claims.
• Built around runtime semantics: ALLOW / VETO and fail-closed UNKNOWN ⇒ VETO.

✅ What it proves (and what it doesn’t)

Proves:

• Behavior at the governed boundary for declared sinks.
• That veto decisions did not trigger irreversible effects through the governed path.

Does not prove:

• Global system security
• Complete correctness of the whole system
• Regulatory certification

📦 Typical Evidence Pack contents (high-level)

• Sink definitions + invariants specification used for governance decisions.
• Test results demonstrating representative ALLOW, VETO, and UNKNOWN / INVALID cases.
• Decision / side-effect correlation artifacts showing:
  • allow ⇒ effect
  • veto ⇒ no effect

🔎 How teams use it

• Engineering: review invariants, validate tests, understand boundary semantics.
• Platform / operations: confirm what was made non-executable by design.
• Internal audit / assurance: validate that controls exist at execution time, not post-hoc.

🧭 When you receive it

• Tier 1: evidence plan and pack structure (definition-level)
• Tier 2: partial pack aligned to scaffold and test harness
• Tier 3: full pack tied to the integrated Customer-Nyxi variant
• Tier 4: refreshed packs as sinks and invariants evolve