The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill contains logic to process external data, specifically Amazon Macie finding details and S3 bucket metadata.
Ingestion points: The scripts/agent.py and scripts/process.py scripts ingest finding data via list_findings and get_findings. Additionally, the Lambda remediation template in SKILL.md processes event details containing finding types and bucket information.
Boundary markers: No specific delimiters or instructions to ignore embedded content within the finding data are present in the scripts or templates.
Capability inventory: The system possesses capabilities to tag S3 objects (put_object_tagging) and publish messages to SNS topics (sns.publish) based on the ingested data.
Sanitization: Finding attributes such as sensitive_types and severity are used directly in S3 tags and SNS messages without explicit validation or escaping.

implementing-aws-macie-for-data-classification