post-quantum-migration-attack

Fail

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: HIGHPRIVILEGE_ESCALATIONDATA_EXFILTRATIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PRIVILEGE_ESCALATION]: The skill instructs the agent to use sudo for high-privilege activities, including searching protected system-wide directories for private keys and performing low-level hardware attacks.
  • Evidence: Execution of sudo grep -rnE "RSA_|ECC_|ecdh_|ecdsa_|BEGIN RSA|BEGIN EC" /etc/ /opt/ /var/ in payloads.md to harvest system credentials.
  • Evidence: Use of sudo ./dramoscope for performing RowHammer attacks on physical memory in payloads.md.
  • Evidence: Use of sudo strings /var/run/qkd/keys.bin to access raw quantum keys from disk in payloads.md.
  • [DATA_EXFILTRATION]: The skill includes extensive patterns for discovering and harvesting sensitive cryptographic materials and high-value data from the host system and network traffic.
  • Evidence: Recursive searches across /etc, /opt, and /var specifically targeting private keys and certificates.
  • Evidence: Instructions for the long-term passive capture of network traffic (TLS handshakes) using tcpdump and tshark for retrospective decryption.
  • Evidence: Python scripts designed to recursively scan /data to identify and classify sensitive data types such as state secrets, genomic data, and PII.
  • [REMOTE_CODE_EXECUTION]: The skill directs the agent to download, build, and install software from external sources at runtime.
  • Evidence: Detailed build instructions for liboqs and oqs-provider requiring git clone, cmake, and make from GitHub repositories.
  • Evidence: Remote script execution logic via frida for process instrumentation and hooking of the Signal messenger protocol.
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the installation of multiple third-party dependencies from public registries.
  • Evidence: Use of pip install for packages including simulaqron, qiskit, cirq, and chipwhisperer in both payloads.md and guides/post-quantum-migration-attack-playbook.md.
  • [COMMAND_EXECUTION]: The skill relies heavily on Python's subprocess and shell commands to manage cryptographic tools such as openssl, ssh-audit, testssl.sh, and ipsec.
  • [DYNAMIC_EXECUTION]: The skill utilizes dynamic process manipulation and instrumentation techniques to audit security protocols.
  • Evidence: Use of frida to hook org.thoughtcrime.securesms and log Signal protocol messages in payloads.md.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from network packets, local filesystems, and remote API responses, creating an attack surface for indirect prompt injection.
  • Ingestion points: Processing tshark network logs, scanning arbitrary files in /data, and ingesting JSON responses from external Signal prekey bundle APIs.
  • Capability inventory: The skill has high-privilege shell access, network request capabilities via curl/requests, and the ability to write to the filesystem.
  • Boundary markers: No explicit delimiters or instructions are provided to the agent to treat external content as untrusted data.
  • Sanitization: There is no evidence of input validation or sanitization when processing data from external ingestion points.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 19, 2026, 04:36 PM