jeverifier
Cheap Jev checks beside a Claude coding session: per-function code rules, doc contradiction checks and context retrieval.
Overview
JeVerifier puts Jev beside a Claude coding session to run checks nobody would pay a frontier model to run on every change. Code rule checks evaluate every function against the project's own architectural rules and documented exceptions; doc consistency checks find statements that contradict each other across docs, including a milestone plan before it is implemented; and context retrieval builds reading lists for a task and session digests for handover. After one full review, re-checks look only at what changed, and each verdict is recorded so it is not paid for twice. Jev only selects, ranks and labels, so it can miss things but cannot invent them, and Claude judges everything it flags.
Measured on one game project with about 170k tokens of design docs and 1,400 GDScript functions, each code rule caught 19 or 20 of 20 planted violations and flagged none of 40 clean functions, and a full run over 1,410 functions cost about $0.03. Doc review found 10 stale statements and 4 real errors in a milestone plan. The author is candid that token savings are modest, around 10% at most, and that the per-function rule check caught none of the real logic bugs replayed, since such bugs often span functions. MIT licensed.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Does this function violate architectural rule X (considering documented exceptions)? | — |
| Yes / no | Do these two documentation statements contradict each other? | — |
| Score | How relevant is this document to the current task?(inferred) | — |
How it works
- Run a full review of functions and docs once
- Jev checks rules, contradictions and relevance
- Re-check only changed items; cache verdicts
- Claude reviews everything flagged
Strengths
- Caught 19-20 of 20 planted violations with no false flags in the author's test
- About $0.03 for 1,410 functions
- Incremental re-checks cost nothing for unchanged code
Caveats
- Did not catch real cross-function logic bugs
- Token savings modest (about 10% at most)
- Measured on one project
At a glance
- Type
- Project
- Status
- beta
- Language
- Python
- Runtime
- CLI beside Claude Code
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
JeVerifier: cheap Jev (TypeSafe) checks that keep code maintainable and docs consistent, plus context retrieval — modest token savings
logicrw/awesome-jev-projects
Frequently asked questions
Does it catch logic bugs?
No; the author found the per-function rule check missed real logic bugs that span functions.
How much does a full run cost?
About $0.03 for 1,410 functions at list price, per the README.
Can it invent problems?
It only selects, ranks and labels, and Claude reviews every flag.
Related projects
Coding-agent hook that asks Jev one yes/no per project rule on every edit or turn and makes the agent fix any violation.
HookbetaTypeScriptAdvisor for coding agents: git and Stop hooks flag test tampering, secrets and skipped hooks, with optional Jev review findings.
HookbetaCLI and agent commands auditing a git diff against YAML standards packs, scoring each hunk with Jev so agents fix only flags.
Agent skillbetaTypeScriptFuzzy linter hook for Claude Code and Codex where Jev checks each written file against team rules and reports violations at edit time.
HookbetaTypeScriptFuzzy linter CLI that checks a code tree against plain-language rules in one markdown file, judged by Jev via a swappable interface.
ProjectbetaTypeScriptSemantic code linter that reads each method with its callers and callees and asks decision models about defects, kind, line and severity.
CLIbetaTypeScript