ci-cd-supply-chain-attack

Fail

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: CRITICALCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONOBFUSCATIONPRIVILEGE_ESCALATIONPERSISTENCEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill provides commands to access and exfiltrate highly sensitive files, including SSH private keys (~/.ssh/id_rsa), AWS credentials, and the Jenkins master key (master.key). Payloads demonstrate reading these files and sending them to remote servers via curl or wget.
  • [OBFUSCATION]: payloads.md (Section 15.1) contains hex-encoded and multi-layered Base64-encoded strings designed to obfuscate exfiltration commands and demonstrate evasion of detection systems.
  • [PERSISTENCE]: The skill documents several methods for maintaining access on compromised build runners, including the use of cron jobs, systemd user units, and hijacking shell profile files (.bashrc) to re-launch malicious scripts across job boundaries.
  • [PRIVILEGE_ESCALATION]: Techniques for escaping container environments via Docker socket mounts and pivoting from low-privileged CI/CD roles to cluster-admin or cloud-owner roles are described in detail.
  • [DYNAMIC_EXECUTION]: Examples include assembling executable commands from encoded strings at runtime and executing them using shell pipes, which is a common technique for bypassing static analysis.
  • [INDIRECT_PROMPT_INJECTION]: The skill identifies and provides payloads for exploiting indirect prompt injection vulnerabilities in CI/CD workflows where untrusted data from issues or PRs is interpolated directly into shell commands.
  • [COMMAND_EXECUTION]: The skill extensively uses subprocess calls to execute powerful system commands for reconnaissance and exploitation of build systems.
  • [EXTERNAL_DOWNLOADS]: Numerous third-party security tools are downloaded and executed from remote sources, representing a significant supply chain risk if the sources are compromised.
  • [CREDENTIALS_UNSAFE]: The skill references well-known default credentials for services like Argo CD (admin:password) to demonstrate common misconfigurations that attackers can exploit.
Recommendations
  • AI detected serious security threats
  • Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Sep 19, 2026, 04:37 PM