affaan-m/agentshield is a security scanning tool that inspects the configuration and permissions of AI agents. It sweeps agent configurations including Claude Code configuration, skills, and hooks to find incorrect permission settings and injection risks. The repository description states that the tool is provided as a CLI, GitHub Action, ECC plugin, and GitHub App integration. The latest release is v1.6.0.
MCP servers (a standard protocol connecting agents with external tools) are also among the inspection targets. According to the README, it detects hardcoded secrets, permission misconfigurations, hook injection, MCP server risks, and agent prompt injection. Findings are organized into A–F grades and a 0–100 score security report, and hardcoded secrets can be automatically fixed to environment variable references. It is written in TypeScript, released under the MIT license, and the repository was created on February 11, 2026. It currently has 1306 stars and a momentum score of 2.7 out of 10.
However, it is not possible to tell from the remaining records what v1.6.0 changed. There are no release change entries, and neither the criteria for the A–F grades nor the formula for the 0–100 score are given. Apart from the star count and the momentum score, there is no metric to gauge performance. It appears to be a tool usable for auditing agent configurations and checking permissions, but the basis on which risk grades are assigned cannot be known until it is checked directly.