verdict

Documentation

Overview

Package verdict is the shared adversarial-verdict value type + evidence bar used by BOTH finding (exploitation) and judgment (AI analysis). It is pure (stdlib + shared only), so neither aggregate imports the other for the one shared evidence-gating mechanism (R1): a distinct verifier seals a verdict whose score moves a claim's standing, and the same EvidenceThreshold gates publishability everywhere – one bar, never forked.

Index

• Constants
• func MeetsBar(score int) bool
• func SelfConfirm(verifier, proposedBy string) bool
• type Verdict
  • func (v Verdict) Validate() error

Constants

const DeterministicProofScore = 90

DeterministicProofScore is the fixed evidence score for a judgment confirmed by a DETERMINISTIC tool proof (a call-graph reachability result), as opposed to a human verdict or an LLM claim. It is well above EvidenceThreshold because a reproducible static call-path is strong evidence, but NOT 100: the underlying analysis (e.g. govulncheck's VTA call graph) is over-approximate, so a "reachable" proof is high-confidence rather than infallible. This is the EVIDENCE/verdict score (gates Publishable); it is distinct from ReachabilityClaim.Confidence (the claim's own self-reported confidence field).

const EvidenceThreshold = 75

EvidenceThreshold is the minimum score (0..100) for a claim-based finding or judgment to be promoted/published. Shared by finding + judgment; do not fork it.

const MaxRationaleBytes = 4096

MaxRationaleBytes bounds sealed verdict rationale payloads and promotion-event provenance.

Functions

func MeetsBar

func MeetsBar(score int) bool

MeetsBar reports whether score clears the evidence threshold.

func SelfConfirm

func SelfConfirm(verifier, proposedBy string) bool

SelfConfirm reports whether a verdict would let a claim confirm itself – the verifier being the same actor that proposed it. A self-confirming verdict must be refused: a finding/judgment cannot confirm itself. An empty proposer (human/tool-sourced, no AI claim) never self-confirms.

Types

type Verdict

type Verdict struct {
	Verifier  string
	Score     int
	Rationale string
}

Verdict is the outcome of an adversarial "try to refute" pass. Score is the verifier's confidence (0..100) that the claim SURVIVES refutation; it becomes the subject's evidence score. Verifier is recorded for provenance and MUST be a DISTINCT actor from the proposer (see SelfConfirm). Rationale is part of the sealed record.

func (Verdict) Validate

func (v Verdict) Validate() error

Validate checks a verdict is well-formed: a recorded verifier, a 0..100 score, and a rationale.