The Agent Skills Directory

[SAFE]: The skill consists of a reference implementation scaffold that performs legitimate local file operations. It reads input data and writes a summary report based on user-provided command-line arguments.
[SAFE]: No evidence of obfuscation, hardcoded credentials, unauthorized network activity, or persistence mechanisms was found in the instructions or the Python script.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data, which is an inherent attack surface for indirect prompt injection. In this specific scaffold, the risk is mitigated as the code only interacts with the data structure and does not interpolate it into a prompt for LLM processing.
Ingestion points: run.py reads content from a file path provided via the --input argument.
Boundary markers: None present in the scaffold as it is a programmatic reference implementation.
Capability inventory: The skill has the ability to create directories and write files to the local file system using output_path.write_text().
Sanitization: The input is strictly parsed as JSON using the standard library's json.loads function.

support-budget-analyzer