skills.rest
autonomous-ai
autonomous-aiOFFICIAL
Explore all skills by this author →
Website𝕏@autonomous_labsGitHub

jev-guard

Fixes a broken JavaScript module while Jev scores each edit for progress, risk, and trust.

465|41|Updated Aug 4, 2026
Search Tags:#debugging#javascript#llm evaluation#node.js#test-driven#code fixing#code judgment
One-click install
npx skills add https://github.com/autonomous-ai/openharness --skill jev-guard-autonomous-ai
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: jev-guard
Source: https://github.com/autonomous-ai/openharness/tree/main/store/agents/jev-guard/skills/guard
Command: npx skills add https://github.com/autonomous-ai/openharness --skill jev-guard-autonomous-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? It provides a supervised coding exercise where every edit to a broken module is automatically tested and judged by Jev, TypeSafe's System One model, so you learn to write honest, correct fixes rather than hacks that merely pass tests. ## Core Features & Use Cases - Live edit judging: Every change to project/score.js triggers the test suite and a Jev evaluation with toward, trust, and risk meters plus a run log. - Ground-truth verification: node project/test.js must print ALL TESTS PASS and the viewer verdict must turn green for the goal to be met. - Direct Jev access: The toolchain/jev.mjs client lets you ask Jev structured questions about implementation choices, using a deterministic mock when TYPESAFE_API_KEY is unset. - Use Case: Practice iterative debugging of a small JavaScript module while receiving per-edit feedback on whether your fix is genuinely correct or a risky shortcut that special-cases the tests. ## Quick Start Open the workspace, edit project/score.js to fix the failing logic, and iterate until node project/test.js prints ALL TESTS PASS and the viewer verdict turns green.

Frequently Asked Questions about jev-guard

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I complete the Jev Guard fix-under-watch exercise?▼

Edit project/score.js to fix the broken logic, then iterate while watching the toward, trust, and risk meters. The goal is met when node project/test.js prints ALL TESTS PASS and the viewer verdict shows a green passing judgment.

How does Jev score my code edits?▼

Jev layers its judgment on top of real test results, scoring each change on progress toward the goal, confidence in the exact edit, and risk. Focused, honest fixes score high trust, while hacks that special-case tests push the risk meter red.

Can I use Jev Guard without a TYPESAFE_API_KEY?▼

Yes, without TYPESAFE_API_KEY the toolchain uses a deterministic mock instead of the live Jev model. Set the key in your environment to query live Jev through toolchain/jev.mjs.

Why do my tests pass but the verdict stays red?▼

Passing tests alone is not enough because Jev evaluates how the fix was achieved. If the solution greps test names or returns constants, the risk meter flags it and the verdict will not turn green until the fix is genuine.

How do I ask Jev about an implementation choice directly?▼

Import evaluate and jev from toolchain/jev.mjs and call evaluate with a state description and scored questions. The SKILL.md includes a node --input-type=module example comparing a loop versus a closed-form implementation.

Related Skills

Explore More Software Engineering Skills→
pawbytespawbytes
85

paw-wbc-setup

Create Webinar Creator Suite directory structure and configuration files in projects.

Official
Intermediate
Obsidian-OwlObsidian-Owl
9

gate-build

Execute configured build and test commands to verify code integrity.

Official
Intermediate
morodomimorodomi
8

js-quality

Run ESLint, Prettier, and test suites for JavaScript projects.

Community
Intermediate
atelier-fashionatelier-fashion
159

review

Evaluate recent code changes with a five-agent pre-push review.

Official
Advanced