The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/netlist_compare.py performs dynamic evaluation of numeric expressions using the eval() function. Although the input is restricted by a regular expression to basic mathematical operators and numbers, the execution of code parsed from external files is a dynamic execution risk.
[COMMAND_EXECUTION]: The skill's scripts, including scripts/spice_order_stress.py and scripts/spice_manifest.py, use subprocess.run() to invoke external circuit simulators. Specifically, spice_order_stress.py allows the simulator command to be configured via a template, which could be exploited for command injection if an attacker influences the command string.
[INDIRECT_PROMPT_INJECTION]: The skill processes SPICE netlist and library files, creating an indirect prompt injection surface.
Ingestion points: The scripts scripts/netlist_compare.py and scripts/spice_order_stress.py read .cir and .lib files.
Boundary markers: No specific delimiters are used to separate circuit data from potential instructions, though the parser is restrictive.
Capability inventory: The skill can read/write files and execute external shell commands via subprocess.
Sanitization: The scripts use a fail-closed parser and a whitelist of safe simulator commands to mitigate risks from malicious netlist directives.
[DYNAMIC_EXECUTION]: The scripts/spice_order_stress.py script generates modified versions of SPICE decks by reordering statements and then executes these files using a sub-process, representing a script-generation-and-execution pattern.

spice-circuit-modeling