The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches pricing data from official AWS Pricing API endpoints (pricing.us-east-1.amazonaws.com and pricing.cn-northwest-1.amazonaws.com.cn). These are trusted domains belonging to AWS. The script fetch-aws-pricing.py includes an explicit check to ensure only HTTPS URLs are accessed, preventing potential URI scheme attacks.
[COMMAND_EXECUTION]: The skill includes Python scripts (fetch-aws-pricing.py, generate-pricing-excel.py) intended for infrastructure analysis. These scripts process command-line arguments using standard libraries and do not involve arbitrary shell execution or unsafe interpolation of external data.
[DATA_EXFILTRATION]: No evidence of unauthorized data exfiltration was found. The network operations are limited to fetching pricing information from AWS, and the output reports (Excel and Markdown) are stored locally in the cost-estimates/ directory.
[CREDENTIALS_UNSAFE]: No hardcoded credentials or secrets were detected. The skill instructions properly guide the user to configure AWS credentials via standard methods (e.g., aws configure) when access to the Pricing API is required.
[INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface by scanning project CDK code and documentation for resource definitions.
Ingestion points: The scanner reads CDK source files (lib/*.ts, infra/*.ts) and architecture documentation (README.md, ARCHITECTURE.md) to identify AWS resources.
Boundary markers: No explicit delimiters are used during the grep-based scan process.
Capability inventory: The skill possesses the ability to write to the local filesystem (generating reports) and perform network GET requests (querying pricing data).
Sanitization: The extraction logic uses strict regex/grep patterns to identify resources, and extracted parameters are used as filters for string matching against pricing data rather than being executed as code. The risk is assessed as low.

cost-estimator