The Agent Skills Directory

[SAFE]: The skill demonstrates legitimate usage of the AWS SDK for Java 2.x to perform cryptographic operations. It adheres to security best practices by utilizing the DefaultCredentialsProvider for credential management and implementing sensitive memory cleanup using Arrays.fill to clear plaintext keys.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the decryption of data from external sources, which creates a generic attack surface for indirect prompt injection. If an agent is instructed to process decrypted content as instructions, it could lead to unauthorized actions.
Ingestion points: Data is ingested through the decryptData and decryptWithEnvelope methods across several files, including SKILL.md and references/technical-guide.md.
Boundary markers: The provided implementation patterns do not include explicit delimiters or warnings to treat decrypted data as untrusted content.
Capability inventory: The skill defines allowed-tools including Bash, Write, and Edit, which represent a significant capability tier that could be targeted by instructions hidden in decrypted data.
Sanitization: The patterns do not include specific sanitization or validation logic for the resulting plaintext data.

aws-sdk-java-v2-kms