The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill mentions using sudo DevToolsSecurity -enable in SKILL.md to configure macOS developer tools, which is a standard procedure to grant debugging permissions to the terminal.
[DYNAMIC_EXECUTION]: Instructions in SKILL.md and references/commands.md describe using lldb's built-in Python interpreter via the script command and executing batch command files using the -s flag for automation.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external data from binaries and process memory, which represents an indirect prompt injection attack surface.
Ingestion points: The agent is instructed to read data from target binaries (target create), process memory (memory read), and symbols (image lookup) as defined in SKILL.md and references/commands.md.
Boundary markers: No specific delimiters or instruction-ignore warnings are specified when the agent processes output from these debugger commands.
Capability inventory: The skill utilizes capabilities such as memory writing (memory write), file export (memory read -o), and arbitrary script execution (script, lldb -s) across the toolset.
Sanitization: The instructions do not include methods for sanitizing or validating data retrieved from target processes.
[COMMAND_EXECUTION]: The skill guides the user and agent through executing various system commands and debugger operations, including lldb, codesign, and package installations via brew or apt.

re-lldb