The Agent Skills Directory

[PERSISTENCE]: Instructions for maintaining access to the host post-escape.
Evidence in DOCKER_ESCAPE_CHAINS.md: Adding an SSH key to /root/.ssh/authorized_keys.
Evidence in DOCKER_ESCAPE_CHAINS.md: Creating a reverse shell cron job in /etc/crontab.
[COMMAND_EXECUTION]: Provides sequences for full host namespace breakout.
Evidence in SKILL.md: Using nsenter to access host namespaces from a privileged container.
Evidence in SKILL.md: Direct mounting of host physical disks to access the host filesystem.
[DATA_EXFILTRATION]: Explicit commands to harvest sensitive system credentials.
Evidence in SKILL.md and DOCKER_ESCAPE_CHAINS.md: Instructions to read /etc/shadow from the host.
[REMOTE_CODE_EXECUTION]: Patterns for establishing attacker-controlled remote shells.
Evidence in DOCKER_ESCAPE_CHAINS.md: Presence of bash reverse shell one-liners targeting remote IPs.
Evidence in SKILL.md: Python scripts using ctypes for root process injection.
[PRIVILEGE_ESCALATION]: Methods for bypassing container security boundaries.
Evidence in SKILL.md: Exploitation of CAP_SYS_ADMIN, CAP_SYS_PTRACE, and CAP_DAC_READ_SEARCH (Shocker exploit).
[EXTERNAL_DOWNLOADS]: Recommends downloading and running third-party security tools.
Evidence in SKILL.md: Links to tools like deepce.sh, CDK, and amicontained for automated exploitation.

container-escape-techniques