The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the execution of the rea-agents CLI tool for system diagnostics (doctor) and setup. It also utilizes system debuggers like lldb and gdb to inspect running processes and core dumps.
[DYNAMIC_EXECUTION]: The observe_native_calls tool executes local binaries under a debugger to trace runtime behavior. While the skill warns that the target may change files or access the network, this represents a significant execution risk. Additionally, the rea-agents setup --yes command is designed to modify the agent's MCP configuration and can replace skill files dynamically.
[EXTERNAL_DOWNLOADS]: The instructions refer to multiple external, third-party repositories and tools for specific analysis tasks (e.g., JADX MCP JAR, Ghidra extensions, and IDA Pro plugins) which are not from verified or trusted vendors.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process and decompile untrusted artifacts (binaries, APKs, JavaScript). The agent interprets strings and pseudocode recovered from these targets, creating a surface for indirect prompt injection where malicious content in the target could influence agent behavior.
[DATA_EXFILTRATION]: While the skill claims to exclude credentials from network captures (inspect_web_network_capture), it provides tools to inspect network traffic and browser/Electron runtime data, which could potentially expose sensitive information if filters are bypassed or misconfigured.

reverse-engineer-anything