AIShield MCP/Agent Security Scan

Version updated for https://github.com/lm203688/aishield to version v4.2.2.

  • This action is used across all versions by 0 repositories.

Action Type

This is a Docker action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action scans and evaluates agents to ensure they adhere to security standards, detecting potential vulnerabilities such as injection attacks and data leaks. It also provides a comprehensive identity system for agents, including DID-based authentication and reputation management. The action integrates with existing toolchains and supports multiple programming languages for safe execution of scripts.

What’s Changed

AIShield v4.2.2

重点：GitHub Action 正式上架就绪

  • 新增 action_entrypoint.py + Dockerfile ENTRYPOINT：Action 现在真正执行扫描（此前只启动服务器），产出 score / risk_level / JSON 报告 / SARIF。
  • action.yml：source_url 改为可选（留空扫当前仓库），默认 no-spawn（不执行被扫配置）。
  • 用法：uses: lm203688/aishield@v4.2.2，fail_on: high，SARIF 直传 GitHub Security。

生态位占位（docs/）

  • docs/ecosystem-positioning-2026.md：agent 生态 4 层 + 数据层盘点，已介入/缺口、开源参考、执行清单。
  • docs/trust-attestation-spec.md：可嵌入发现层的中立信任凭证 aishield-trust/v1（Server Card / Agent Card / ai-catalog 用 trust 字段引用）。
  • docs/.well-known/agent-card.json 已 dogfood trust 字段。
  • distribution/guardrail-harness/：forge / Goose / Open Interpreter 的 drop-in 内容安全护栏示例。

规则与能力

  • 214 MCP / 220 Skill 规则，OWASP MCP Top 10 + Agentic ASI01–10 双维对齐。
  • Agent 计算机「内容安全平面」叙事（与隔离运行时互补）。

Marketplace 发布需在仓库 Marketplace 标签页点击 Publish（需已验证的发布者身份）。