The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of instructions attempting to bypass safety filters, override system prompts, or extract internal configuration was found. The instructional content is strictly focused on MDVRP optimization.
[DATA_EXFILTRATION]: No sensitive file paths, hardcoded credentials, or unauthorized network operations were detected. The skill performs local mathematical computations.
[REMOTE_CODE_EXECUTION]: There are no patterns involving the download and execution of remote scripts (e.g., curl | bash). All code is provided statically within the skill.
[COMMAND_EXECUTION]: The Python scripts do not use subprocess, os.system, or any other method to execute arbitrary shell commands.
[EXTERNAL_DOWNLOADS]: No external URLs or remote resources are referenced for download or execution. All dependencies are standard Python packages.
[OBFUSCATION]: The code and documentation are provided in clear text with no evidence of Base64 encoding, zero-width characters, homoglyphs, or hidden instructions.
[DYNAMIC_EXECUTION]: No use of eval(), exec(), or other dynamic code generation techniques was observed. The skill uses standard function calls and object-oriented patterns.
[PRIVILEGE_ESCALATION]: No commands for elevating privileges (e.g., sudo) or modifying system configurations were found.
[PERSISTENCE_MECHANISMS]: The skill does not attempt to modify startup scripts, cron jobs, or registry keys to maintain persistence.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied coordinates and demands for optimization. While this is an ingestion surface, the code handles data as numeric types for mathematical processing and does not interpolate untrusted strings into agent prompts, effectively neutralizing this risk.

multi-depot-vrp