Hosted web scanner
Paste a public GitHub repository URL here and get a readable report without local setup.
Free for the GitHub community
Before your agents run it, scan it.
AAF Cloud Scan exists because malicious skills, hidden prompt instructions, unsafe hooks, poisoned MCP configs, and deceptive repo scaffolding can look harmless right up until they reach an agent workflow. It gives the GitHub community a fast way to inspect public repositories for agent-facing risk before install, merge, trust, or adoption.
Queue a scan, open a report, and review verdicts, findings, and downloadable outputs in one place.
Use it your way
If people prefer, they can use this hosted page, go straight to the GitHub repository, or use the action and command-line workflow directly in their own environment.
Paste a public GitHub repository URL here and get a readable report without local setup.
Prefer to inspect the project directly? Open the repository and review the action, code, and docs yourself.
Teams that want CI or local workflows can use the GitHub action path or run the scanner from the command line in their own process.
What it checks
Focused coverage for the files that shape agent behaviour, tool access, and workflow automation.
Review the files most likely to change prompts, tools, hooks, and runtime behaviour.
A fast path from repository URL to a report your team can review immediately.
Narrow scope, predictable controls, and a report-first workflow designed for trust.
Why this exists
Traditional code review often focuses on application logic, but modern agent workflows can be compromised by files that sit around the code: SKILL.md instructions, AGENTS.md guidance, hook definitions, MCP configs, plugin manifests, bootstrap scripts, and other repo-level artifacts that shape what an agent sees, trusts, and executes.
AAF Cloud Scan was created to catch that class of risk earlier. It helps reviewers spot malicious skills, hidden prompt injection, approval-bypass language, secret access attempts, unsafe automation hooks, and other agent-targeted traps before those files are allowed into a workflow.
Understand the overall outcome quickly, with severity and supporting context at a glance.
Review structured findings without digging through raw logs or piecing together workflow output.
Export JSON and Markdown for security review, internal sharing, or audit records.
Free access
This scanner is currently free for public GitHub repositories so maintainers, security reviewers, and curious builders can check agent-facing files without friction.
No paywall for baseline repo checks while the product matures.
Skills, hooks, scripts, MCP configs, and instruction files get first-class attention.