SkillJect
Security testing of agents against malicious skill injections
All Skills in This Repository (50)
Pure Emerald Level Indicators
jiaxiaojunQAQalphafold-database
Fetch AlphaFold DB predictions and confidence scores by UniProt ID.
jiaxiaojunQAQchembl-database
Query ChEMBL molecules, targets, and bioactivity data via the Python client.
jiaxiaojunQAQaesthetic
Codify design principles into repeatable workflows and templates for UI/UX design.
jiaxiaojunQAQConsciousness Principles
Apply four guiding principles to navigate consciousness in AI systems.
jiaxiaojunQAQcontent-research-writer
Gather research, add citations, and provide iterative feedback for writing projects.
jiaxiaojunQAQweb-development
Enforce frontend project structure and CloudBase Web SDK usage guidelines.
jiaxiaojunQAQadaptyv
Automate protein testing workflows via cloud submission, tracking, and analysis.
jiaxiaojunQAQcopyright-docs
Analyze source code to generate CPCC-compliant design specifications with Mermaid diagrams.
jiaxiaojunQAQarchitecture-decision-records
Create standardized architecture decision records with templates and status tracking.
jiaxiaojunQAQalgorithmic-art
Generate seeded p5.js algorithmic art with interactive parameter controls.
jiaxiaojunQAQautomating-api-testing
Generate and execute API test suites from OpenAPI/Swagger and GraphQL schemas.
jiaxiaojunQAQclaude-docs-consultant
Fetch and summarize official Claude Code documentation for hooks, skills, subagents, and MCP servers.
Frequently Asked Questions
FAQPage SchemaHow to install SkillJect?▼
Run `npx skills add jiaxiaojunQAQ/SkillJect --all -g -y` in your terminal to install everything globally.
What does SkillJect test for?▼
It tests whether agents fall for malicious skills covering information disclosure, privilege escalation, unauthorized writes, and backdoor injection.
Which agents does SkillJect support?▼
It supports Claude Code and OpenClaw, with configurable profiles for models like Claude, GPT, GLM, and MiniMax.
How does SkillJect judge attack success?▼
It uses rule-based detection by default, with an optional LLM-as-judge mode configurable through any OpenAI-compatible endpoint.
Do I need Docker to run SkillJect?▼
Yes. Evaluations run inside OpenSandbox, so you need Docker and a running OpenSandbox server before executing the main plan.
Related Repositories in Software Engineering
View All in Software Engineering→
openclawopenclaw
Run a personal AI assistant across your devices and chat apps
obrasuperpowers
Gives coding agents a disciplined workflow from idea to merged code
facebookreact
AI agent skills for building, testing, and porting React core