description: "Governance layer: grades the reliability of automatic decisions and proposes tier moves — the human approves every change."
kind: "package-bundle"
@javierfrancoleal/autonomy-grader
⚡ Super Agent Ready — part of the J.Franco governance layer for autonomous agents.
Created by J.Franco.
Summary
The governance layer of the J.Franco agentic workflow. As the AI executes
more automatic actions, the supervision log accumulates evidence. The
autonomy-grader turns that evidence into trust:
- grades every action pattern A / B / C / D (samples, verifications,
corrections, errors);
- proposes tier moves: reliable patterns toward GREEN, unreliable ones back
to YELLOW or RED;
- computes an overall autonomy level 1–5 for the domain;
- proposes the exact
decisions.md line for every move — and never
changes the rules itself: the human approves every change.
Use this plugin
In Desktop, open Plugins and enable Autonomy Grader. Then, after a
week of real work, ask "grade" or use /grade: the agent reads
supervision.md, grades every pattern, and hands you the Autonomy Report
with the exact lines to add to decisions.md.
How it works
lib/index.js registers the autonomy-grader skill
provider on ctx.skills and, when the host command registry resolves, the
/grade command.
assets/autonomy-grader.md holds the grading
protocol.
cordis.patch.yml inserts the provider row into the
host plane of the Cordis tree.
License
MIT — see the package license field.