ARIS
Governed agent workflows with mission lifecycle checks, per-call authorization, and local governance history.
01 / The engineering question
What is this project for?
Make agent execution conditional on explicit mission state, approval policy, and agent membership.
- Inputs
- Persisted mission metadata, allowed agent names, and an execution identity. Raw prompts are outside the authorization decision.
- Outputs
- Mission state, governance events and rejection evidence. These are local software records, not a certified audit service.
02 / Documented workflow
How the work is structured
- 01
Create a mission and define its permitted agents.
- 02
Request and record approval when the mission policy requires it.
- 03
Check lifecycle, approval and exact membership before each correlated agent call.
- 04
Record governance decisions and recover interrupted local persistence.
03 / Source material
Evidence you can inspect.
The public README documents mission lifecycle transitions, approval prerequisites, mission-correlated agent authorization, and event-first persistence on a local POSIX filesystem.
Read the primary evidence ↗- Authorization policyaris/core/authorization.py
- Mission lifecycle modelaris/core/mission.py
- Public authorization tests (not executed here)tests/test_authorization.py
Source links are pinned to reviewed commit 64cef726af4d. Later repository changes do not silently change this evidence.
04 / Verification boundary
What this evidence does not prove.
Documentation establishes design and interfaces; this website has not independently executed or certified ARIS. Raw prompts and successful outputs can remain in its local ledger. It is not a distributed authorization service.
Source inspection is not an execution test, a security certification, a performance benchmark, or proof of a production deployment.
05 / Next verifiable step
Build the next piece of proof.
Publish a reproducible synthetic demonstration of approval, denial, quarantine, and recovery.