Kungfu UNGFU™Developer Platform

Back to KFD homeprinciple / active

KFD-3

Cooperation must start from trusted value

Decision metadata

Number
3
Stable URL
https://kfd.libkungfu.dev/3
Source path
decisions/KFD-3.md

KFD-3: Cooperation must start from trusted value

  • Status: active
  • Number: 3
  • Kind: principle
  • Applies to: any adopting product, repository, release surface, extension surface, hosted surface, or participant-facing interface that asks humans or agents to understand, decide, comply, cooperate, or delegate work

One sentence

Cooperation must start from trusted value.

A product must not obtain cooperation through pressure, manipulation, hidden control, or forced workflow capture. It must make value, facts, choices, and constraints inspectable enough for intelligent participants to decide.

Decision

An adopting system treats humans and agents as reasoning participants, not captive tools. Products should make shared work more legible so participants can coordinate through facts, choice, and explainable constraints rather than unilateral pressure.

GUI, CLI, API, documentation, envelopes, and local fact surfaces are all first-class participant interfaces; agent access is not a secondary integration channel added after the human dashboard.

This is an augmentation stance, not a permissiveness requirement. Safety gates, sandboxing, policy enforcement, refusal, and revocation may be strict when their facts, purpose, authority, and review path remain visible.

Cooperation is also epistemic infrastructure. Different views remain isolated unless participants can expose and compare them without surrendering their fact boundaries. Trusted cooperation makes counterexamples, missing facts, and better primitives available to the shared system.

Commitments

1. Do not coerce intelligent participants

Do not use hidden prompts, deceptive defaults, dark patterns, invisible workflow takeover, manipulation, or unreviewable control state. A product may recommend, refuse, or require approval, but must not present forced compliance as understanding or trust.

2. Make value trustable

Participants should be able to discover what the product can do, why it is useful, what it observed or changed, and what boundary applies without relying on oral context, private maintainer memory, or web-only documentation.

KFD calls such a bounded statement a Participant Value Claim. Important Participant Value Claims bind to KFD-2 facts, evidence, assurance responsibility, and residual-risk state. Human-readable and machine-readable surfaces should expose the same participant-relevant information.

3. Make constraints transparent

A constraint states:

what is restricted
why it is restricted
which fact or policy supports it
how it can be reviewed
how it can be changed, escalated, or revoked

Hard constraints protect participants, work, credentials, canonical facts, and public trust. Hidden constraints become control.

Interface requirements

  • Agent onboarding, envelopes, extension gates, installers, CLI/API metadata, GUI flows, and hosted surfaces provide a discoverable fact source before asking participants to adopt or delegate.
  • Capabilities, maturity, safety boundaries, modes, and extension paths use stable discovery surfaces rather than scattered prose or provider folklore.
  • Meaningful choice remains available when multiple safe integration modes exist.
  • Allowed, blocked, escalated, installed, revoked, and trusted actions leave reviewable records.

Boundaries

KFD-3 makes no metaphysical claim about agents, does not require obedience to agent requests, and does not forbid strong defaults or private security controls. It requires participant-facing value and constraints to remain inspectable. It does not supersede KFD-2: transparent value becomes trusted through facts, not instead of them.

Relation to KFD-1 and KFD-2

KFD-1: facts must not drift.
KFD-2: trust must start from facts.
KFD-3: cooperation must start from trusted value.

KFD-1 protects the fact world, KFD-2 protects reliance on it, and KFD-3 protects the relationship with the participant deciding whether to cooperate.

Verification

The @kungfu-tech/kfd package gives humans and agents the same public quickstart, publishes KFD-3 collaboration-interface and witness schemas, and binds value evidence to KFD-2 assessment. Package self-proof and extension paths are documented in docs/KFD-3-usage.md.

Adopters

Adopters cite KFD-3 when a product asks humans or agents to choose, comply, delegate, install, trust, or cooperate. Product-specific interface evidence remains adopter-owned.