dl-binary-recompilation-engineer
Warn
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructions in SKILL.md direct the agent to execute 'npx rea-agents setup' to configure the environment if an MCP server is missing. This command fetches and runs code from the npm registry, introducing a risk of executing unvetted third-party scripts at runtime.\n- [EXTERNAL_DOWNLOADS]: The skill refers to an external GitHub repository (https://github.com/morluto/rea) for binary analysis capabilities. This repository is not maintained by a recognized trusted organization or the skill author, potentially exposing the environment to unverified software.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted binary files and metadata derived from those binaries. This creates a vulnerability surface where malicious content embedded in binaries (e.g., in strings or section names) could influence the agent's behavior.\n
- Ingestion points: Reads local native binaries and outputs from tools like Ghidra and REA.\n
- Boundary markers: Missing specific instructions to treat decompiled output or binary strings as untrusted or to use delimiters.\n
- Capability inventory: The skill has significant capabilities including code generation, compiler invocation, and shell command execution.\n
- Sanitization: No sanitization logic is specified for the data extracted from binary files.\n- [DYNAMIC_EXECUTION]: The primary purpose of the skill is to generate, compile, and potentially execute native code (recompilation). While this is the intended functionality, the process involves dynamic generation and execution of code based on external binary input.
Audit Metadata