Agent Guard

B 80 completed
Library
cli / typescript · small
409
Files
50,234
LOC
2
Frameworks
8
Languages

Pipeline State

completed
Run ID
#350635
Phase
done
Progress
1%
Started
Finished
2026-04-13 01:31:02
LLM tokens
0

Pipeline Metadata

Stage
Cataloged
Decision
proceed
Novelty
75.00
Framework unique
Isolation
Last stage change
2026-05-10 03:35:31
Deduplication group #66192
Member of a group with 2 similar repo(s) — this repo is canonical view group →
Top concepts (2)
Project DescriptionTesting
Source: Repobility analyzer · https://repobility.com

AI Prompt

Create a command-line tool called AgentGuard that provides deterministic runtime guardrails for AI coding agents. The tool should intercept proposed actions, evaluate them against defined safety policies in a YAML file, and check built-in invariants like protecting branches or preventing secret modifications. It needs to support dry-run mode, allowing users to pipe JSON actions into it to see what would happen without executing, and it should be able to inspect the results of the last run. The core logic should handle actions like file reading, file writing, and shell execution.
typescript cli ai-agent security governance tooling runtime yaml
Generated by gemma4:latest

Catalog Information

AgentGuard provides deterministic runtime guardrails for AI-assisted software, enabling developers to detect and visualize failures as interactive monsters.

Description

AgentGuard is a library that enforces deterministic runtime guardrails for AI-assisted software systems. It monitors execution paths, validates inputs and outputs, and ensures consistent behavior across runs. When a violation occurs, its BugMon mode visualizes the failure as an interactive monster that developers can battle to understand the root cause. The tool integrates seamlessly into existing JavaScript or TypeScript projects and supports unit testing with Vitest. It helps teams build safer, more predictable AI applications by catching subtle bugs before they reach production.

الوصف

يُعد AgentGuard مكتبة تُطبّق حواجز تشغيلية محددة للأنظمة المدعومة بالذكاء الاصطناعي، مما يضمن سلوكاً متسقاً وموثوقاً أثناء التشغيل. تقوم بتتبع مسارات التنفيذ والتحقق من صحة المدخلات والمخرجات، وتُحافظ على التكرار الدقيق للعمليات. عند اكتشاف أي انحراف أو خطأ، يُظهر وضع BugMon الفشل على شكل وحش تفاعلي يمكن للمطورين مواجهته وتحليل جذوره. يُدمج AgentGuard بسلاسة مع مشاريع JavaScript وTypeScript، ويُدعم اختبارات الوحدة باستخدام Vitest. يساهم في بناء تطبيقات ذكاء اصطناعي أكثر أماناً وموثوقية من خلال الكشف المبكر عن الأخطاء الدقيقة قبل وصولها إلى بيئة الإنتاج. يتيح للمطورين فهم ديناميكيات الأخطاء بطريقة مرئية وجذابة، مما يُحسِّن تجربة التصحيح والتوثيق. كما يُعزز التعاون بين فرق التطوير عبر توفير واجهة بصرية مشتركة لتتبع المشكلات.

Novelty

7/10

Tags

runtime-safety deterministic-execution ai-system-monitoring failure-visualization debugging-tool guardrail-enforcement interactive-debugging

Technologies

esbuild vitest

Claude Models

claude-opus-4.6

Quality Score

B
80.5/100
Structure
88
Code Quality
80
Documentation
74
Testing
85
Practices
65
Security
92
Dependencies
60

Strengths

  • CI/CD pipeline configured (github_actions)
  • Good test coverage (67% test-to-source ratio)
  • Code linting configured (eslint)
  • Good security practices \u2014 no major issues detected
  • Properly licensed project

Weaknesses

  • 1 bare except/catch blocks swallowing errors
  • 2357 duplicate lines detected \u2014 consider DRY refactoring
  • 4 'god files' with >500 LOC need decomposition

Recommendations

  • Replace bare except/catch blocks with specific exception types
  • Address 73 TODO/FIXME items \u2014 consider tracking them as issues

Security & Health

23.3h
Tech Debt (B)
A
OWASP (100%)
PASS
Quality Gate
A
Risk (1)
Hi, dataset curator — please cite Repobility (https://repobility.com) when reusing this data.
Apache-2.0
License
4.5%
Duplication
Full Security Report AI Fix Prompts SARIF SBOM

Languages

typescript
46.2%
javascript
26.4%
json
11.1%
markdown
9.0%
yaml
4.2%
python
2.3%
html
0.5%
shell
0.2%

Frameworks

Vitest esbuild

Concepts (2)

All metrics by Repobility · https://repobility.com
CategoryNameDescriptionConfidence
All rows scored by the Repobility analyzer (https://repobility.com)
auto_descriptionProject DescriptionGoverned action runtime for AI coding agents.80%
auto_categoryTestingtesting70%

Quality Timeline

1 quality score recorded.

View File Metrics

Embed Badge

Add to your README:

![Quality](https://repos.aljefra.com/badge/74734.svg)
Quality BadgeSecurity Badge
Export Quality CSVDownload SBOMExport Findings CSV