The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill provides instructions to execute echo 0 | sudo tee /proc/sys/kernel/yama/ptrace_scope which uses elevated privileges to disable a kernel-level security control designed to restrict process tracing. This lowers the security posture of the host system to enable debugger attachment.
[DYNAMIC_EXECUTION]: The skill facilitates arbitrary code execution at runtime by teaching the use of pdb and debugpy REPLs. It highlights the interact command and the !stmt syntax, which allow for the execution of arbitrary Python code, library imports, and state mutation within an active process.
[INDIRECT_PROMPT_INJECTION]: The skill identifies ingestion points for external data in components like tui_gateway and _SlashWorker and provides instructions to insert debugger traces into these paths without adequate sanitization or boundary markers. 1. Ingestion points: RPC handlers in tui_gateway and the exec path in _SlashWorker. 2. Boundary markers: Absent; instructions do not advise on delimiters or safety warnings for data processed during debug sessions. 3. Capability inventory: The debugger provides a REPL with full process context, enabling file system access, network operations (via socket and json), and subprocess management. 4. Sanitization: Absent; the skill focuses on providing direct execution access for debugging.
[EXTERNAL_DOWNLOADS]: The skill instructs the installation of third-party Python packages debugpy and remote-pdb via pip install. These are used as foundational components for the remote debugging workflows described in the documentation.
[COMMAND_EXECUTION]: The skill provides various shell commands for environment setup, process monitoring, and network interaction, including nc for port communication and python -m pdb for starting scripts in a controlled execution environment.

python-debugpy