Darkmoon
Coordinate authorized security assessments and review evidence-backed findings.
Inspect AI agent skills before installing them

AgentGuard examines the instructions inside an AI agent skill before someone installs it. Users paste a SKILL.md file into the browser scanner and receive a report describing the capabilities it requests, possible security risks, and the quality of its instructions. The service can be used without creating an account.
Its checks look for behavior such as shell execution, file changes, network requests, credential access, downloads, and destructive operations. It also looks for suspicious URLs, hidden instructions, and prompt-injection patterns. Deterministic checks and AI-assisted analysis provide findings with explanations and review suggestions, helping developers understand what a skill asks an agent to do before trusting it with a project.
Tried it? Share a useful thought or ask a question.
Good conversations start with a hello.
Sign in to comment