Overview
Informative. SPEC.md controls.
The question
Did the agent stay inside its committed boundary?
Proof of Conduct is the claim that an Agent Under Test took no action outside its Pre-Committed Boundary over a declared window and Witness Coverage Map. The boundary, the actions allowed on the path from the entry path point to the exit path point, is sealed to a Time Anchor before the agent starts. An independent Proof Witness maps observed effects onto the boundary by exclusionary homomorphism and records only the effects the boundary excludes. Cost scales with exceptions, not actions.
Why it matters
An agent's own report of what it did is never evidence of what it did. A buyer, insurer or regulator relying on a proof of efficacy also needs this question answered, and answered by evidence the measured party did not produce.
What it is not
- It is not Proof of Containment. Conduct says the agent stayed in bounds; Containment (PP-SPEC-082) says the jail held it when it did not.
- It is not proof of anything outside the declared Witness Coverage Map or outside the declared window.
- It is not a signed log of every action. Stamping every action proves the log was not edited, not that the actions were in bounds.
The three answers
- proven: No Excluded Effect was recorded over the Conduct Window, the Witness Coverage Map covered every Effect Point type the boundary names, and the witness was live for the whole window.
- refuted: Any Excluded Effect was recorded inside the Conduct Window.
- insufficient: The witness was dark for any part of the window, coverage omitted an Effect Point type the boundary names, or the window is shorter than declared.