Can Gitingest users test a no-write graph proof against a prompt digest?

[nike-17]

Disclosure: I maintain Sverklo.

Gitingest is a strong baseline for prompt-friendly repo context. I am testing a different first-run shape: a no-write command that prints central files, one selected symbol, callers, and a prompt for a coding agent before writing MCP config.

I am trying to collect corrections and grep-better cases, not stars.

From a repo root:

npm exec --yes --package=sverklo@latest -- sverklo prove --no-write --guided --markdown

Good feedback would be one of:

• a repo where a digest is more useful than the graph proof
• a wrong center/symbol/caller result
• generated or example files dominating the receipt
• setup friction
• a case where grep/ripgrep is simply better

Public proof thread: sverklo/sverklo#79

Proof sprint tracker: sverklo/sverklo#85