The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches tools and configuration from several well-known repositories within the reverse engineering community.
Evidence: Provides instructions to clone repositories from organizations like radareorg, rizinorg, and frida, as well as established researchers such as mrexodia, LaurieWired, and cryptax.
[REMOTE_CODE_EXECUTION]: The installation process for several MCP bridges involves downloading source code from GitHub and installing it locally using pip install -e .
Evidence: Installation instructions for ida-pro-mcp and radare2-mcp involve cloning the source and executing the local setup script.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze, disassemble, and decompile untrusted binary files, which presents a surface for potential indirect prompt injection through strings, function names, or metadata contained within the binaries.
Ingestion points: Binary files and executable code provided for analysis in SKILL.md.
Boundary markers: The skill does not define specific delimiters to separate untrusted binary data from agent instructions.
Capability inventory: The skill has capabilities to read files, run analysis tools, and modify tool state (e.g., renaming functions, adding comments) across Ghidra, IDA, and radare2.
Sanitization: No explicit filtering or sanitization of binary metadata is described before it is processed by the agent.
[COMMAND_EXECUTION]: The skill utilizes a variety of command-line tools for binary analysis, system package management, and debugger operations.
Evidence: Use of radare2, git, pip, brew, and apt commands for environment setup and binary manipulation.

reverse-engineering