audit

name	audit
description	Find -> adversarially verify -> rank issues over a target (a diff, dir, file, or glob). Use for code review, a scoped bug hunt, a security or perf pass. Run via the workflow tool by name.
kind	workflow
source	builtin

audit — find, verify, rank

A workflow preset. Run it with the workflow tool:

workflow(name="audit", args={"target": "apps/server/ntrp/server", "dimensions": ["bugs", "security"], "depth": "normal"})

Parallel finders surface issues along each dimension, every fresh finding is adversarially verified by independent skeptics, and only the survivors are ranked into a report. Loops until a round finds nothing new (capped by depth).

args

• target (str, default ".") — what to examine: a dir, file, glob, or "the diff".
• dimensions (list[str], optional) — e.g. ["bugs", "security", "performance"]; defaults to a correctness/security/perf/edge-cases set.
• depth ("quick" | "normal" | "deep", default "normal") — scales skeptics-per-finding and max rounds.
• finder_model / synth_model (str, optional) — override the model for finders / the final synthesis; omit to inherit the run's model.