The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the dap binary from the author's GitHub repository (AlmogBaku/debug-skill) as part of the setup process.
[EXTERNAL_DOWNLOADS]: The skill fetches the js-debug adapter from Microsoft's official GitHub repository (microsoft/vscode-js-debug) to enable Node.js debugging.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute dap CLI commands for controlling program execution, inspecting memory, and evaluating expressions.
[PERSISTENCE]: The installation script scripts/install-dap.sh suggests adding the tool's installation directory to shell configuration files (e.g., ~/.bashrc, ~/.zshrc) to maintain the tool's availability in the PATH environment variable across sessions.
[PRIVILEGE_ESCALATION]: The documentation references the potential need for elevated privileges, such as running sudo DevToolsSecurity -enable on macOS or attaching to processes via PID, which are standard operational requirements for debugger functionality.
[INDIRECT_PROMPT_INJECTION]: The skill processes live program data, such as variable values and stack traces, which are captured as output from the dap tool and could potentially contain untrusted strings from the application being debugged.
Ingestion points: Tool output from dap eval, dap inspect, and automatic context updates at breakpoints.
Boundary markers: None explicitly implemented to isolate program-generated values from debugger metadata.
Capability inventory: The skill possesses capabilities to read files, inspect process memory, and execute shell commands through the dap interface.
Sanitization: No specific sanitization or filtering of the debugger's output is performed before it is provided to the agent.

debugging-code