Reproducing the operational-assurance experiment

This guide is the bounded experimental implementation closeout for #96 and its children #98–#100. It does not document a live two-registry pilot, current authorization proof, production execution, or third-party certification.

Reproduce from a clean checkout

Use the repository’s documented Python, Node and Rust dependencies. From the repository root run:

make setup
make validate
make typescript-check
make rust-check
make cross-runtime
make network-interop
make pages-check

Alternatively make release-check-all runs the governed release checks. The command suite depends on installed runtimes and the repository’s toolchain guidance; the checks are also performed by GitHub Actions. Do not infer that this document itself executes the commands or proves third-party reproduction.

Bounded evidence entry points are conformance/test-vectors/operational-assessment/v0.1.0.json, conformance/test-vectors/operational-dependency/v0.1.0.json, conformance/test-vectors/operational-reliance/v0.1.0.json, and conformance/test-vectors/operational-receipt-binding/v0.1.0.json. Output includes artifacts/operational-assessment/validation-report.json, artifacts/operational-dependency/validation-report.json and artifacts/operational-reliance/. The current TypeScript comparison is composition-only: its observation and dependency dimensions originate in Python.

Reproduction boundaries

flowchart TD
 A["Modeled observations and dependency declarations"] --> B["Python assessment evidence"]
 B --> C["Independent TypeScript composition comparison"]
 C --> D{"Outcomes and reasons agree?"}
 D -->|No| E["Regression: fail"]
 D -->|Yes| F["Bounded modeled equivalence"]
 G["Current authoritative agent and delegation state"] -. "Not proven by fixtures" .-> F
 H["Real cross-registry recognition and verified receipt"] -. "Requires independent evidence" .-> F
 F --> I["No authority implication or production certification"]

IR-01, IR-07 and IR-09 in #100 have modeled analogues but cannot be represented as independently verified live identity, federation or A2A invocation tests. Receipt reference consistency is not cryptographic validation. A privately disclosed observation is schema-valid but access-control enforcement is not established by the model.

Acceptance disposition

For the original #96 tranche, the user authorized a bounded closure scope: accept modeled optional experimental implementation separately from production and independently verified assurance. The original checkboxes remain historical requirements; unmet requirements must not be checked merely because they were transferred. The two independent obligations have explicit successor tracking:

  • #113: independent observation provenance, privacy and cross-registry recognition.
  • #114: real decision and execution receipt assurance.

This is not a normative specification amendment. Closing #96 and its children on an accepted bounded scope does not close #113 or #114 and does not claim that their acceptance criteria passed.

Release disposition

The new profiles are informative experimental artifacts and do not change Candidate v0.10.0 or IETF draft -04. Release decision is separate and must follow docs/release-policy.md and the checked source commit, with compatibility and limitations stated explicitly; no release/version is implied by issue closure.


Agent Registry Protocol documentation. Specification text and documentation are licensed as stated in the repository NOTICE and license files.

This site uses Just the Docs, a documentation theme for Jekyll.