The Agent Skills Directory

[SAFE]: Technical analysis of the core scripts (debt_scanner.py, debt_prioritizer.py, and debt_dashboard.py) reveals they are legitimate utilities using standard Python libraries for file parsing (AST) and reporting. No malicious patterns, persistence mechanisms, or unauthorized privilege escalation were detected.
[CREDENTIALS_UNSAFE]: The assets/sample_codebase/ directory contains files with intentional hardcoded credentials (e.g., sk-1234567890abcdef in user_service.py and sk_test_EXAMPLE_NOT_REAL in payment_processor.py). These are explicitly documented as mock data for demonstration and training purposes for the scanner tool and do not represent a vulnerability in the tool's execution.
[PROMPT_INJECTION]: A surface for indirect prompt injection exists in scripts/debt_scanner.py as it processes external codebase comments.
Ingestion points: Reads the content of every file in the user-specified directory (e.g., lines 173-195).
Boundary markers: Extracted code patterns and comments are placed directly into JSON output fields without delimitation or instructions to the agent to ignore embedded content.
Capability inventory: The tool only performs static analysis and reporting; it does not execute the code it processes.
Sanitization: No sanitization is performed on the extracted text before it is written to the report. However, since the tool is for manual review and the agent does not execute output, the risk is negligible.
[COMMAND_EXECUTION]: The scanner tool uses the ast module to safely analyze Python code structure and re for other languages. It does not use eval(), exec(), or subprocess to execute untrusted code or shell commands.

tech-debt-tracker