The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs extremely forceful and coercive language to override the AI's normal operational logic and safety guardrails. Instructions such as "ABSOLUTELY MUST invoke," "YOU DO NOT HAVE A CHOICE," "This is not negotiable," and "not optional" are used to compel the agent to load sub-skills even with a "1% chance" of applicability. The skill further includes a "Red Flags" section that explicitly commands the agent to suppress its own reasoning and caution (e.g., "STOP—you're rationalizing").
[COMMAND_EXECUTION]: The skill directs the agent to execute multiple bash scripts located in a directory path computed from the $SUPERHACKERS_ROOT environment variable. Scripts such as detect-tools.sh, preflight.sh, run-tool.sh, engagement-init.sh, and validate-output.sh are invoked using the bash shell. Because these scripts are external to the skill and their contents are not provided, they represent an unverified execution path.
[COMMAND_EXECUTION]: In the "Error Classification" table, the skill provides an explicit instruction to "Run with sudo" when encountering "Permission denied" or "Operation not permitted" errors. This constitutes a direct instruction for the agent to attempt privilege escalation on the host system.
[COMMAND_EXECUTION]: The skill encourages the use of a wide array of powerful security and exploitation tools, including nmap, sqlmap, msfconsole, msfvenom, and bettercap. It provides specific execution patterns such as backgrounding processes (&) and setting timeouts, which are typical of automated attack workflows.
[REMOTE_CODE_EXECUTION]: The reliance on external scripts defined by environment variables ($SUPERHACKERS_ROOT) and the instruction to download/reference files like SETUP.md and TOOLCHAIN.md from external locations (referenced via relative paths likely pointing to a larger repository) creates a surface for executing code not contained within the audited skill.

using-superhackers