feature idea, check the quote still matches the block

[Flyingmiata-droid]

Vault Operator is the only agent plugin I have found that puts block ids on the claims at ingest, so this is a gap question, not a wishlist.

The block id tells me which paragraph a claim came from. It does not tell me the words still match it. That breaks two ways. I edit a source note later and the claim in my sense making note quietly stops matching, while the link still resolves fine. Or the agent tightens a passage into something quote shaped at ingest, and nothing re-reads the block to check.

I built a small plugin for that one job, Cite Engine, live in the community directory. No AI in the check itself. It snapshots the quoted text, re-reads the source block, and flags anything that drifted or no longer resolves.

Is a fidelity check like that worth having inside Vault Operator, or would you rather point people at it?

[pssah4]

Thanks for the issue and idea, I'll have a look.

[Flyingmiata-droid]

Thanks for the issue and idea, I'll have a look.

Appreciate it. To keep the look quick, here is the shape of it.

The check is deterministic and self contained, no model call, so it adds no AI surface to Vault Operator. It reads a block by id, snapshots the quoted text, then re reads that same block later to see if the words still match or the id stopped resolving.

Since you already write block ids at ingest, most of the plumbing is already there. Would you want it running as a check inside Vault Operator, or is it cleaner for you to just point people at the plugin?