Hi everyone,
I’m building Atlas, a metadata-first proof/evidence system for critical digital actions.
The basic idea is simple: important actions like CI releases, approvals, AI-agent events, production-readiness checks, and privileged workflow changes should come with a receipt.
Atlas does not try to replace GitHub, CI/CD, scanners, approval tools, or business systems. Instead, it records and verifies the proof chain around them:
who requested the action
what capability or policy applied
whether approval was required
what evidence or artifact references were emitted
what commit, packet, or receipt contains the result
how another reviewer can replay or verify the proof
what the proof does not show
The goal is to reduce evidence-chasing and make reviews clearer without storing raw secrets, customer data, private logs, prompts, packet captures, or sensitive artifacts.
I’m looking for feedback on:
whether the proof-receipt model makes sense
whether the reviewer flow is understandable
where the docs feel confusing
what risks or blind spots I should explain better
what teams would need before trying this in a real workflow
A good starting point is the README and reviewer quickstart.
Thanks for taking a look. Any feedback helps.