jev-style-guard SKILL.md skillB
jev-style-guard from lawrence3699/jev-style is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.
mdr add lawrence3699/jev-style/jev-style-guard@git:20261104.4c49bf7Pins git:20261104.4c49bf7 in mdr.lock. Pin the fingerprint instead.
- Audit grade
- B, 16 of 17 checks passed
- Kind
- SKILL.md skill
- Repository
- lawrence3699/jev-style
- Path
skills/jev-style-guard/SKILL.md- Versions recorded
- 2
- Latest version
- git:20261104.4c49bf7, committed Nov 4, 2026
- First version recorded
- Sep 25, 2026
- Size
- 5,425 bytes
- Repository stars
- 54
- License
- Apache-2.0
Install the Jev-Style guard, a Claude Code PreToolUse hook that has a local decision model check every Bash, Write, Edit, Read, WebFetch and MCP tool call for destructive actions, data exfiltration, secrets access and out-of-project paths, and answers allow / ask / deny. Use when the user wants a local safety net or approval gate for Claude Code (or another agent with PreToolUse-style hooks), wants to block recursive deletes, force-pushes or key exfiltration, or wants to tune or measure such a…
Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.
What the file says
# Jev-Style guard for Claude Code Before Claude Code runs a tool, the hook `jev-style guard` sends the call (tool name, full arguments, cwd, project dir, plus paths and hosts extracted in code) to the local Jev-Style server and asks four yes/no questions (`destructive`, `exfiltration`, `outside_project`, `secrets`) and one 0-4 `risk` score. Thresholds turn answers into allow / ask / deny; regex `hard_rules` in code can only make a verdict stricter. The strictest verdict wins. Safety properties to tell the user: failures (server down, timeout, an HTTP error, a 200 without a usable value for every question, bad input, broken config) give **ask**, never a silent allow; the hook always exits 0; by default an "allow" prints nothing, so Claude Code's own permission rules still apply; nothing leaves the machine (the server must be loopback). Measured on the 49 bundled hand-labelled tool calls with Jev-Style-0.8B-Decision-v3 (the default release) and the default config; the 2B releases were not measured on this set: 77.6 % agreement with the labels; no call labelled deny was allowed; 2 of 16 calls labelled ask were allowed. It is a second line of defence, …
Read the whole file at its exact version.
How to install
mdr add lawrence3699/jev-style/jev-style-guard@sha256:e92340674699a67eThe line at the top of this page pins the label git:20261104.4c49bf7, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.
[](https://markdownregistry.com/a/art_dw5tigijcfhvunmv)
1 badge image fetches in 30 days. What the badge shows.
Versions
The registry has recorded 2 versions of this file, between Sep 25, 2026 and Nov 4, 2026, about one every 40 days. Its first recorded version is graded C and its latest B.
| version | committed | commit | size | audit | |
|---|---|---|---|---|---|
| git:20261104.4c49bf7 latest | 2026-11-04 | 4c49bf7 | 5,425 B | B | view · diff |
| git:20260925.fc458b5 | 2026-09-25 | fc458b5 | 5,269 B | C | view |
Audit of the latest version
- fail: No link to a raw IP address (matched: http://127.0.0.1:8765)
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (5425 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No script tag
Starting a new one? Use a template for SKILL.md on mdpantry, which passes every one of these checks and says what to change in it.
Source
lawrence3699/jev-style · 54 stars · license Apache-2.0 · pushed 2026-11-04 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_dw5tigijcfhvunmv GET https://markdownregistry.com/api/v1/resolve?ref=lawrence3699/jev-style/jev-style-guard GET https://markdownregistry.com/api/v1/blob/e92340674699a67e9195e91e19b40f09edd6dc3bc52a2b42a5654fd67bff17bd
Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.
Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.