jev-bugfix
Codex skill using the local Jev CLI to rank candidate code snippets before reproduction, root-cause analysis and a verified fix.
Overview
jev-bugfix is a Codex skill, documented in Chinese, for bounded Jev-assisted bug investigation. Codex reproduces the failure, collects short contiguous snippets from the stack trace, rg searches and call relationships, reviews the full input, optionally asks Jev once to score each candidate's relevance, and then confirms the root cause, makes a minimal fix and verifies it with the original reproduction and related tests. Jev only supplies relevance scores, a Python helper produces the investigation order, and Codex does the reasoning; simple problems with obvious causes can be fixed directly.
All candidates are kept, stack-trace-indicated snippets are investigated first, and if scoring fails investigation continues locally. Candidates are prepared by hand as JSON, reviewed for secrets with sensitive snippets marked local only. A fixed-pool comparator benchmarks stack-first, lexical and Jev orderings offline (coverage, Hit@1/3, MRR, simulated bytes read), and a run summariser keeps failures and cost completeness. The README states there is not yet any conclusion about real reading, time or cost savings. It has offline tests and an MIT licence; the Jev CLI is a separate dependency.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Score | How relevant is this code snippet to the observed bug? | — |
How it works
- Reproduce the failure
- Collect candidate snippets from stack, rg and call relationships
- Review inputs for secrets; optionally score once with Jev
- Investigate in the produced order; confirm root cause
- Make a minimal fix and verify with tests
Strengths
- Keeps every candidate; stack-indicated ones first
- Works offline if scoring fails
- Offline benchmark tools to compare orderings
Caveats
- No measured savings yet, per the README
- Candidates are assembled manually
- Documentation in Chinese
Getting started
`python3 -B skills/jev-bugfix/scripts/rank_candidates.py --input tests/fixtures/smoke_case.json` for an offline dry run; reference SKILL.md from Codex.At a glance
- Type
- Agent skill
- Status
- prototype
- Language
- Python
- Runtime
- Codex skill
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
A Codex skill for bounded Jev-assisted bug investigation, root-cause analysis, and verified fixes.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev do in bug fixing?
It scores candidate snippets for relevance to set the investigation order.
Does it save time?
The README says there is no conclusion on savings yet.
What if Jev is unavailable?
Investigation continues locally.
Related projects
Maps each Python function to the tests that run it; Jev scores candidate function-test links that static analysis misses.
ProjectbetaPythonCLI that scores every test against a git diff with Jev and emits filter arguments for vitest, jest, node, Playwright, cargo, go and rspec.
ProjectbetaTypeScriptMCP server that returns only the JS/TS code a task needs by following the TypeScript compiler graph and asking Jev about relevance.
MCP serverbetaTypeScriptSemantic grep for code, diffs and CSV rows that packs 16 chunks per Jev request and prints file:line hits with CI-friendly exit codes.
Agent skillbetaTypeScriptLocal code search for coding agents: ripgrep/BM25 shortlist, then Jev judges each chunk's relevance, served over MCP.
MCP serverbetaRust- jevgrep (dzhng)★ 2,671
Code search CLI and agent skill (jg) where Jev judges relevance across folders, files and declarations to return excerpts.
Agent skillbetaTypeScript