The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The script scripts/refresh.sh is designed to clone the repository https://github.com/trimstray/the-book-of-secret-knowledge.git and overwrite internal skill reference files with the downloaded content.
[COMMAND_EXECUTION]: The skill provides a vast collection of powerful and potentially dangerous shell commands in references/shell-oneliners.md and references/shell-tricks.md. Notable examples include bind shells (nc -l 5000 -e /bin/bash), reverse shells using named pipes, and commands for modifying /etc/profile or truncating system files.
[REMOTE_CODE_EXECUTION]: The self-update mechanism in scripts/refresh.sh presents a supply-chain risk. Because the skill fetches and extracts text from an external 3rd-party source to populate its own reference files, a compromise of that remote source could result in the delivery of malicious instructions that an agent might subsequently recommend or attempt to execute.
[DATA_EXFILTRATION]: The functions DomainResolve and GetASN in references/shell-tricks.md use curl to transmit hostnames and IP addresses to external services (dns.google.com and ip-api.com). While intended for diagnostics, this patterns represents a data exposure surface.
[PROMPT_INJECTION]: The SKILL.md file contains instructions in the 'Operational Gotchas' section that direct the agent to use specific execution methods like /bin/bash -lc or a Python ThreadPool for parallel operations, which functions as a low-level attempt to override default agent execution behavior.

sysadmin-toolbox