The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The runtime requirements in SKILL.md instruct the user to configure the agent to use npx -y binary-ninja-mcp, which downloads the binary-ninja-mcp package from the npm registry at runtime.
[REMOTE_CODE_EXECUTION]: By using npx, the skill facilitates the execution of remote code from a third-party package repository.
[COMMAND_EXECUTION]: The skill configuration relies on executing shell commands (e.g., npx -y binary-ninja-mcp --host localhost --port 9009) to interface with the Binary Ninja analysis environment.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted binary files, which is a known vector for indirect prompt injection if malicious data is embedded in the binary's metadata or strings.
Ingestion points: Binary files are loaded into the analysis context via the load_binary tool (mcp-tools.md).
Boundary markers: The instructions do not define specific delimiters or warnings to separate untrusted binary data from the agent's control logic.
Capability inventory: The skill includes tools to read raw bytes (read_bytes) and perform write operations like renaming functions (rename_function) and setting comments (set_comment).
Sanitization: There is no evidence of sanitization or validation logic for the data extracted from the binary files before it is processed by the agent.
[DATA_EXFILTRATION]: The load_binary and read_bytes tools allow the agent to access and read data from absolute file paths on the local system. While this is necessary for binary analysis, it presents a risk of sensitive data exposure if the agent is manipulated into accessing unintended files.

binary-ninja-mcp-analysis