The Agent Skills Directory

[SAFE]: The skill demonstrates strong adherence to security best practices for serverless environments. It explicitly recommends using environment variables for configuration to avoid hardcoded secrets and advocates for modular AWS SDK v3 imports to minimize the function's attack surface and cold start times.
[PROMPT_INJECTION]: The skill's architecture involves ingesting data from external, potentially untrusted sources, creating a surface for indirect prompt injection.
Ingestion points: Data enters the agent's context through event.body, event.pathParameters (API Gateway), and SQS record.body as seen in SKILL.md.
Boundary markers: The provided handler templates do not implement specific boundary markers or instruction isolation techniques to prevent the LLM from executing commands embedded within the data.
Capability inventory: The skill provides patterns for interacting with AWS services including DynamoDB (GetItem, PutItem), SQS (message processing), and S3 (signed URLs and object streaming).
Sanitization: The examples use standard JSON parsing (JSON.parse and json.loads) but do not include content-level sanitization or filtering to detect malicious prompt instructions.

aws-serverless