kernel-audit-security

The skill is purpose-aligned and not covertly exfiltrative, but it is a high-risk offensive security capability for an AI agent. It directs the agent to pen-test codebases, develop exploit evidence, run delegated auditors, and potentially execute/build target code; that belongs to its stated purpose, yet materially increases misuse and operational risk. Install trust is moderate: source appears to be same-org Cloudflare content via the documented Skills CLI workflow, with no curl|bash or opaque binary forwarding, but also no signed release trail. Overall: suspicious/high-risk capability, not confirmed malware.

SUSPICIOUS: the skill is internally coherent and not a credential harvester or supply-chain lure, but it is an offensive security skill that enables an AI agent to perform penetration-style analysis, run local tests, and autonomously investigate exploit paths. The flagged scanner items are mostly benign documentation artifacts; the main risk is the intentional security-testing capability itself, not hidden exfiltration or installer abuse.

The supplied code demonstrates severe Android access-control and credential-handling vulnerabilities: exported components expose password data, bypass authentication, permit password changes, and disclose decrypted passwords through SMS and logs. These are high-impact security risks. The content is readable and instructional, with no meaningful obfuscation or evidence of conventional malware such as a hidden backdoor, cryptominer, or unrelated exfiltration mechanism; however, the demonstrated receiver intentionally performs unauthorized credential disclosure when invoked.

This is an exploit demonstration rather than ordinary application functionality. It shows intent hijacking and attacker-controlled ContentProvider behavior that can lead to path traversal and arbitrary native library execution when a victim application inadequately validates implicit-intent results and loads content:// or file:// resources. The fragment itself does not show execution against a victim, but the example components are intentionally malicious in their described deployment context.

The fragment documents a WebView bridge exploitation technique that can retrieve and display a secret from a vulnerable iOS application. It presents a security risk to applications that expose sensitive native methods to untrusted WebView content and use unsafe DOM insertion, but it does not contain standalone malware or supply-chain compromise behavior.