The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the installation of the dap CLI tool from a community GitHub repository (AlmogBaku/debug-skill). The provided scripts/install-dap.sh script downloads pre-compiled binaries via curl from an unverified source.
[REMOTE_CODE_EXECUTION]: The skill promotes several remote code execution vectors, including go install for unversioned community packages and a scripted download/extraction process for the js-debug adapter.
[COMMAND_EXECUTION]: Multiple high-risk execution patterns are identified:
Persistence: The scripts/install-dap.sh script modifies user shell configuration files (.bashrc, .zshrc, etc.) to permanently alter the system PATH.
Privilege Escalation: The skill documents and suggests actions requiring elevated privileges, such as writing to system-wide binary directories, enabling DevToolsSecurity, and specifically mentioning the disabling of macOS System Integrity Protection (csrutil disable).
Arbitrary Execution: The core functionality involves attaching to running processes and remote hosts, allowing for arbitrary expression evaluation within the target program's execution context.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. It processes live program data, variable states, and stack traces without sanitization or boundary markers. Malicious content within a debugged application could potentially inject instructions into the agent's context during inspection.

debugging-code