The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill initiates downloads of the 'dap' binary and various debug adapters from external sources.\n
Evidence: Downloads from github.com/AlmogBaku/debug-skill via scripts/install-dap.sh and microsoft/vscode-js-debug via references/installing-debuggers.md.\n- [REMOTE_CODE_EXECUTION]: The installation process fetches pre-compiled binaries from a community GitHub repository and executes them on the host system.\n
Evidence: The script scripts/install-dap.sh uses curl to download a binary to ~/.local/bin or /usr/local/bin and applies chmod +x.\n- [COMMAND_EXECUTION]: The skill uses shell commands to interact with debuggers and manage process execution.\n
Evidence: Uses dap debug, dap step, dap continue, and allows attaching to processes via PID or host:port.\n- [PERSISTENCE]: The installation script suggests modifications to the user's shell profile to ensure the tool remains in the system PATH.\n
Evidence: Recommends adding export statements to ~/.bashrc, ~/.zshrc, or ~/.config/fish/config.fish.\n- [DYNAMIC_EXECUTION]: The skill allows for the evaluation of arbitrary expressions against a running program's state.\n
Evidence: The dap eval "<expr>" command facilitates runtime execution of code within the context of the debugged process.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingest and displays output from external programs and evaluated expressions which could contain malicious instructions.\n
Ingestion points: Program output (dap output), variable inspection (dap inspect), and expression evaluation (dap eval) in SKILL.md.\n
Boundary markers: The skill uses structural headers like 'Locals:', 'Stack:', and 'Output:' to delimit data.\n
Capability inventory: Includes shell command execution, process attachment, and file system modification via installation scripts.\n
Sanitization: No explicit sanitization of debugged program output or evaluation results is documented.

debugging-code