confidential-computing-attack
Audited by Socket on Sep 19, 2026
3 alerts found:
Securityx3This skill is coherent with its stated purpose, but that purpose is to equip an AI agent for offensive security operations against confidential-computing systems. The main issue is not hidden behavior or malware; it is that the skill grants high-risk exploit capability, secret-recovery guidance, and defense-evasion tactics with execution and modification tools suitable for real-world misuse.
This is an offensive confidential-computing attack playbook, not conventional package malware. It contains actionable procedures for reconnaissance, attestation replay and bypass, side-channel exploitation, secret recovery, private-key discovery, governance takeover, manifest tampering, and persistence. There is no evidence that the document itself automatically executes or exfiltrates data, but its misuse could cause severe confidentiality and integrity failures. It should be treated as high-risk security content and restricted to authorized, isolated testing environments.
The fragment is a security assessment/red-team reference containing potentially dangerous operational commands and explicit examples of credential discovery, governance takeover, and data exfiltration. The displayed JavaScript backdoor and CCF commands are high-risk if executed, but the visible material does not demonstrate that the package automatically performs these actions. Treat the source as dual-use offensive-security content and review the complete file, package scripts, and entry points before use.