Route focused code-review investigations

01 · Role in the system

What Jev does here

Jev supplies bounded risk, file, evidence, mechanism, severity, and reviewer-routing judgments across a sequence of focused calls. Code discovers Git changes or repository files, enforces workflow order, persists reports, and serves a local dashboard. The resulting findings are prompts for human review rather than proof that a defect exists.

A review workflow made of six bounded judgments

Code review is usually presented as one big language-model task: read the diff, find the bugs. This workflow decomposes it into a sequence of narrow calls — screen for risk, select evidence, classify the mechanism, score severity, and route the finding to the right kind of reviewer. Each call has a short option list and a typed answer, which is what lets code enforce the order of the workflow and persist a report a human can actually work through.

The output posture matters as much as the pipeline. Findings are described as prompts for human review, never as proof a defect exists, and the tool deliberately stays away from compiler diagnostics, static analyzers, and repository indexing. It is a triage layer over plain Git state, and it says so.

Where this record sits among review tools

Compared with LLM review bots that generate prose comments, this pattern spends its model budget on routing decisions: which files deserve a look, which evidence supports a suspicion, how severe the finding is. Those judgments are checkable in a way generated prose is not — a reader can inspect the selected evidence and disagree with the severity score, but cannot be fooled by fluent writing.

What the record does not establish is precision and recall. No labeled defect set was run through the workflow, so the honest reading is: the mechanism works and produces reviewable reports, while its hit rate against real bugs is unknown. The limitations field states exactly that.

02 · Control boundary

Where Jev sits

Broad risk screening narrows into evidence selection and severity scoring under a code-owned review workflow.

Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.

03 · Known limits

What this evidence does not prove

• It does not integrate compiler diagnostics, static analyzers, or repository indexing.
• Model-selected findings require independent review and do not prove that a defect exists.

04 · Attribution

Public sources

This is a Community record: the project was published by a third-party community author.

• devagrawal09 ↗Community · github · public · checked 2026-09-18