The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to utilize elevated privileges via sudo for several operations, including installing root CA certificates on macOS and running low-level tracing tools like dtruss. These commands provide the agent with high-level access to the host operating system.
[DATA_EXFILTRATION]: In the native binary debugging methodology for macOS, the skill provides specific steps to set up a mitmproxy and trust its certificate system-wide using sudo security add-trusted-cert. This configuration allows the agent to intercept and decrypt all TLS-encrypted system traffic, which represents a critical risk for data exposure if misused.
[EXTERNAL_DOWNLOADS]: The skill requires the installation and execution of numerous external tools and libraries from public registries (NPM, PyPI, Homebrew) and GitHub. These include specialist binaries like Ghidra, Playwright, and various reverse-engineering scripts, which significantly expand the attack surface of the execution environment.
[PROMPT_INJECTION]: The 'Oracle Triple' and 'Debug Squad' workflows (detailed in methodology references) establish a mechanism where untrusted data—such as user-provided bug descriptions, logs, and code—is ingested and passed to automated subagents. This creates a significant surface for Indirect Prompt Injection, as the skill lacks instructions for sanitizing this input or using clear boundary markers to prevent malicious instructions within the data from influencing the agent's or subagents' behavior.
[DATA_EXPOSURE]: The skill provides instructions for capturing and inspecting raw memory, registers, and network payloads which may contain sensitive runtime data, including API keys or environment variables, especially during the reverse-engineering of bundled applications.

debugging