KFD-3 formal reference

Authoritative decision · Formal model · Usage · Documentation map

Status: experimental
Normative: no
Formal model version: 1
Authority: decisions/KFD-3.md
Decision status: active

Imported vocabulary

Participant, ParticipantValueClaim, TrustAssessment, ChoicePath, Constraint, ExtensionPath, CooperationEvent.

Domain objects

For participant p and offered work o:

Value(o, p)        Participant Value Claim relevant to p
Trust(o, p)        KFD-2 assessment visible to p
Choice(o, p)       safe available modes, including decline where applicable
Constraint(o, p)   restriction, purpose, authority, review, and revocation
Interface(o, p)    discoverable participant-facing surface

Define cooperation eligibility:

Eligible(p, o) :=
  Discoverable(Interface(o, p))
  and Inspectable(Value(o, p))
  and Inspectable(Trust(o, p))
  and Visible(Choice(o, p))
  and Transparent(Constraint(o, p))
  and not Coerced(p, o)

Eligibility does not force cooperation. It establishes a valid basis from which the participant may decide.

Relations and predicates

Discoverable(i)    a participant can find the interface from a minimal entry
Trusted(v)         value claim v is bound to a KFD-2 assessment
Transparent(x)     a constraint exposes purpose, fact or policy, authority,
                   review, and change or revocation path
Coerced(p, o)      cooperation is obtained through hidden pressure,
                   manipulation, deceptive default, or unreviewable control
Closed(I)          reachable participant surfaces are classified

Invariants

I1  RequestsCooperation(o, p) -> Inspectable(Value(o, p))
I2  Trusted(Value(o, p)) -> Visible(Trust(o, p))
I3  Enforced(Constraint(o, p)) -> Transparent(Constraint(o, p))
I4  SafeAlternativeExists(o, p) -> Visible(Choice(o, p))
I5  ParticipantFacing(i) -> Discoverable(i) and Classified(i)
I6  ValidCooperation(p, o) -> Eligible(p, o)
I7  SafetyGate(x) does not imply Coercion when Transparent(x)

Cooperation transition

participant discovers interface
  -> value and facts become inspectable
  -> trust and residual risk become visible
  -> choices and constraints become visible
  -> participant chooses, declines, or escalates
  -> action leaves a reviewable record

Proof obligations

Identify every participant profile materially affected.
Provide at least one minimal discovery entrypoint.
Bind important value claims to KFD-2 facts and assessments.
Expose safe choices and transparent constraints.
Classify reachable participant-facing surfaces.
Record allowed, blocked, escalated, installed, revoked, or trusted actions.
Expose an extension request path when the declared interface is insufficient.

Invalid states

Hidden prompts or control state obtain apparent cooperation.
Human documentation exposes participant-relevant information unavailable to an agent interface, or the reverse, without a declared reason.
A callable or participant-facing surface exists outside a claimed closed-world interface.
A safety restriction has no visible purpose, authority, or review path.
A value claim is presented as trusted without a KFD-2 fact boundary.

Machine mappings

Formal statement	Decision source	Schema or check	Verification
Participant, entrypoint, value, choice, constraint	Interface requirements	`schemas/kfd-3/collaboration-interface.schema.json`	Machine for structure
`I5` interface closure	Commitments, Interface requirements	`schemas/kfd-3/witness.schema.json`	Machine for declared boundary
Trusted value binding	Make value trustable	KFD-2 trust assessment links	Mixed
KFD package self-interface	Verification	`.buildchain/kfd-3/`, `scripts/check.mjs`	Mixed

Non-claims and extension points

KFD-3 makes no claim that every participant has identical capability, authority, or access. It does not prohibit refusal, sandboxing, policy enforcement, or revocation. It requires the basis of cooperation and control to remain inspectable. New participant profiles and interface kinds may be added without changing this reference.

KFD-3 formal reference

Formal reference metadata