A little more clarity. A lot less setup.
Welcome to AutoSPECTRA.
Understand where your foundation model generalizes.
Start with a prompt in the agent you already use.
Where does your model generalize?
Give your agent a model and a question. AutoSPECTRA guides the investigation.
Follow protocol.md (https://spectra.yashaektefaie.com/protocol.md) to audit the generalizability of [model name and identifying information]. Once you believe you have finished, fetch and follow finish.md (https://spectra.yashaektefaie.com/finish.md) before giving your final response. Available compute and access: [hardware, runtime limits, data or checkpoint locations, and other deployment information].
Fill in your model and available resources, then paste into your coding agent.
Has someone already tested it?
Check what is known about a model before you put it to work.
Find audits of [model name or paper link] in AutoSPECTRA. Start at https://spectra.yashaektefaie.com/api/models and follow the returned model, axis, and audit-card links. Read all relevant cards in JSON and Markdown, then summarize their findings, evidence, limitations, and links to the original results. Distinguish model versions and tasks, and tell me if no audit matches.
Replace the brackets with the model you are interested in.
Make your findings useful to others.
Have an audit to share? Your agent can prepare it for the registry.
Follow https://spectra.yashaektefaie.com/submit.md to submit my completed audit of [model name] for review using the HTTPS API. My report and supporting files are at [path or link]. Prepare and validate the full submission bundle, submit it, and save the receipt and status-check token privately. Report the returned status; a pending submission is not published.
Give your agent the audit report and supporting files along with this prompt.
Or just give your agent this link.
It can help you choose where to start.