The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data in the form of .NET assemblies and binaries. This creates a potential surface for indirect prompt injection if the agent attempts to follow instructions embedded in the analyzed code or metadata.
Ingestion points: Processes external .exe and .dll files via tools like dnSpyEx and de4dot (SKILL.md, references/obfuscators.md).
Boundary markers: Includes a mandatory confirmation gate requiring explicit user authorization and scope definition before processing (SKILL.md).
Capability inventory: Utilizes subprocess calls for binary analysis, file system writes for patching, and debugging capabilities (SKILL.md, references/common-workflow.md).
Sanitization: Explicitly instructs the user to use isolated, non-persistent virtual machines and block external network access when analyzing unknown samples.
[COMMAND_EXECUTION]: The skill utilizes several powerful command-line utilities and system commands for its primary workflow, including powershell, file, grep, and specialized tools like de4dot and diec.
[EXTERNAL_DOWNLOADS]: The documentation directs users to download and install various reverse engineering tools and MCP extensions from public GitHub repositories and official registries such as Chocolatey and NuGet.
[DYNAMIC_EXECUTION]: Provides guidance on using the dnlib library to programmatically manipulate IL (Intermediate Language), which involves the generation and modification of executable code segments at runtime as part of the deobfuscation or patching process.

dotnet-reverse