Atlas Cloud Generator / guides / editorial
A provenance guide for an Atlas Cloud Generator brief
Atlas Cloud Generator provenance guide gives producers concrete criteria to preserve input-to-output provenance, with verified inputs, explicit limits, decision evidence, and accountable next steps.
Create a traceable cloud-generation chain from authorized references and prompt versions to job records, downloaded outputs, and use approval.
Connect prompts, settings, and jobs
Store prompt versions as text with timestamps and editors rather than screenshots alone. Link each submitted job to the exact prompt, input versions, controls actually selected, account or workspace, operator, submission time, and available service event ID. If the interface omits a field, mark it unavailable instead of reconstructing it from memory. Preserve change reasons between attempts so reviewers can explain why an output differs. The record should distinguish creative revision, source substitution, setting change, service retry, and unknown variation.
Register outputs and downstream use
Assign each returned file an asset ID and retain its original bytes, checksum when appropriate, measured properties, thumbnail, linked job, and review disposition. When an editor crops, composites, retimes, or combines it, create a derivative record instead of overwriting origin. Link final use approval to the precise derivative in the released master. Document deletion or retention actions for cloud inputs and outputs according to policy. This chain proves what your team recorded; it does not certify vendor security, ownership, availability, output quality, or reproducibility beyond the observed job.