The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of prompt injection attempts, safety filter bypasses, or instructions to ignore system guidelines. The content is strictly focused on route optimization and algorithmic problem-solving.
[DATA_EXFILTRATION]: No sensitive file path access, hardcoded credentials, or suspicious network operations were found. The skill operates on local numeric data such as distance matrices and coordinates.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns detected. There are no instances of fetching and executing scripts from external URLs (e.g., curl | bash).
[COMMAND_EXECUTION]: The skill includes Python scripts for algorithmic calculations. These scripts use standard libraries (NumPy, PuLP, OR-Tools) for linear programming and combinatorial optimization. No dangerous subprocess calls or shell command execution on untrusted data were identified.
[EXTERNAL_DOWNLOADS]: Mentions well-known and trusted Python libraries and solvers (Google OR-Tools, PuLP, Gurobi, CPLEX). These are industry-standard tools for optimization. No downloads from untrusted or unknown sources are requested.
[OBFUSCATION]: The file was analyzed for Base64, hex encoding, zero-width characters, and homoglyphs. No obfuscation techniques were found; all code and instructions are in plain text.
[PRIVILEGE_ESCALATION]: No commands for acquiring administrative privileges or modifying system-level configurations were detected.
[PERSISTENCE_MECHANISMS]: No attempts to establish persistence (e.g., modifying shell profiles or cron jobs) were found.
[DYNAMIC_EXECUTION]: No use of unsafe dynamic execution functions like eval(), exec(), or unsafe deserialization (pickle) was observed. The logic is implemented through static algorithmic code.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: Processes user-provided distance matrices and coordinate lists in SKILL.md.
Boundary markers: Not explicitly defined, but inputs are expected to be strictly numerical data types (integers/floats).
Capability inventory: Code performs mathematical calculations and logic for TSP solvers. No file-writing or network operations are included in the scripts.
Sanitization: Standard Python type handling (int/float/np.array) provides inherent validation against non-numeric injection in the context of these specific scripts.
Conclusion: Risk is minimal as the attack surface is limited to numerical inputs for algorithmic solvers.

traveling-salesman-problem