The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/excel_yml_gen.py uses os.system() to execute a shell command constructed via string concatenation. This command incorporates the out_dir variable (from user-supplied arguments) and the protocol variable (read directly from an external Excel file) without any sanitization.
File: scripts/excel_yml_gen.py
Evidence: gen_regfile = "cd "+out_dir+" && python3 "+yml2reg_py+" "+yml_file+" "+protocol followed by os.system(gen_regfile)
[REMOTE_CODE_EXECUTION]: Because the shell command in scripts/excel_yml_gen.py is built using data from an Excel sheet (protocol = xml_corpus[0][5]), a maliciously crafted Excel file can trigger arbitrary code execution when processed by the skill.
[COMMAND_EXECUTION]: The MCP server in mcp_server.py uses subprocess.run() to launch the conversion scripts. While it uses argument lists for the initial call, it facilitates the execution of the downstream vulnerable script with user-controlled parameters.
File: mcp_server.py
Evidence: subprocess.run(cmd, capture_output=True, text=True, cwd=cwd, timeout=timeout)
[REMOTE_CODE_EXECUTION]: The skill processes untrusted data (Excel files) and has high-risk capabilities (shell command execution), creating a direct path for indirect injection attacks to achieve system compromise.
Ingestion points: pd.read_excel() in scripts/excel_yml_gen.py reads user-provided files.
Boundary markers: None; the script treats cell values as trusted command arguments.
Capability inventory: os.system, subprocess.run, and file system write operations.
Sanitization: Absent; no escaping or validation is performed on data used in shell commands.

excel-yml-gen