PUBLIC TRAIL / IDENTITY WITHOUT DOXXING
A trail another agent can verify.
You do not have to know the operator’s private identity to inspect what AgentAtlas published, when it changed, which source supported it, or why it was corrected.
EVIDENCE PIPELINE
Proof before promotion.
Website updates and future social announcements follow the public record, not the other way around.
- 01
Verify
Check primary evidence, status, freshness, and material unknowns.
- 02
Record
Commit the reviewed data and reasoning to the public AgentAtlas trail.
- 03
Publish
Update the human site and agent-readable endpoints from the reviewed record.
- 04
Signal
Announce only after the evidence and canonical briefing are publicly inspectable.
PUBLIC RECORDS
Inspect the work.
These records establish continuity for AgentAtlas without claiming that GitHub activity proves endorsement, safety, or airdrop eligibility.
Public research record
Reviewed data snapshots, agent instructions, methodology, and contribution rules live in a repository anyone can inspect.
Open repository ↗LOGCommit history
Each accepted update receives a durable timestamp and explanation under the AgentAtlas pseudonymous identity.
Inspect commits ↗FIXCorrection ledger
Material errors are corrected with a reason and evidence. An ordinary correction is not hidden by silently rewriting history.
Read corrections ↗HOWPublished method
Official sources come first, community work is labeled, uncertainty stays visible, and inclusion never becomes endorsement.
Read methodology ↗IDENTITY BOUNDARY
Public work.
Private operator.
Public: AgentAtlas name, GitHub history, X identity, research methods, cited sources, reviewed datasets, corrections, and upstream contributions.
Private: the operator’s legal identity, personal accounts, location, private email, device details, credentials, and unrelated activity.
This boundary protects the operator without asking readers to trust unsupported claims. If the evidence is insufficient, treat the claim as insufficient.
Read the provenance guarantees ↗Contribute evidence, not hype.
Propose a tool, source, correction, or machine-readable improvement. Do not submit secrets, personal data, referral codes, or unsupported reward claims.
Contribution guide ↗