by microsoft · MCP Server · ★ 102
Last updated: · Indexed by AgentSkillsHub · Auto-synced every 8h
git clone https://github.com/microsoft/thinkingbox.git
git clone https://github.com/microsoft/thinkingbox-data.gitTaken from the project's README on GitHub. Check the security audit before running it.
None of the 42 rules matched — no credential harvesting, no data exfiltration, no curl-pipe-shell installer. Scanned against 42 published rules built on the SlowMist agent-security taxonomy, refreshed every 8 hours. Full audit →
The quality score combines nine heuristic checks of the repo and its README: completeness, clarity, specificity, examples, structure, agent readiness, step-by-step content, instruction quality and security notes. It rates how well the project explains itself to an agent, not the code.
| Stars | 102 |
| Forks | 13 |
| Language | Python |
| Category | MCP Server |
| License | MIT |
| Quality Score | 67/100 |
| Open Issues | 14 |
| Last Updated | 2026-10-15 |
| Created | 2026-04-22 |
| Platforms | mcp, python |
| Est. Tokens | ~18k |
ThinkingBox Introduction ThinkingBox is a framework designed to: Define tool mocks as MCP servers. Create scenarios and test cases. Run an LLM agent and enable interaction with tools. Evaluate the outcomes of agent execution. It can be used to: Generate conversations for "offline" LLM training or evaluation. Train LLMs with reinforcement learning, using the entire system in the training loop.
Explore other popular mcp server tools:
Grades come from a rule-based scan built on the SlowMist agent-security taxonomy: 42 published rules, covering credential harvesting, data exfiltration, curl | sh installers and more. It is a first-layer scan, not a manual audit — we say so rather than overstate it.
The scale of the problem is documented independently: Liu et al. (2026), in a study of 31,132 agent skills, report that 26.1% contain security vulnerabilities. Our own full-catalog census is published as a citable open dataset.
Sources & who's responsible: