The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes arbitrary string inputs for encryption and decryption, establishing a potential surface for indirect prompt injection if the output is later consumed by an LLM.
Ingestion points: Data enters the skill context through the plainText parameter in the encrypt function and the cipherText parameter in the decrypt function.
Boundary markers: No specific delimiters or instructions (e.g., 'treat this as untrusted data') are suggested in the usage examples to isolate these inputs from the agent's instructions.
Capability inventory: The skill is limited to calling AWS KMS cryptographic APIs; it does not demonstrate capabilities for shell execution, subprocess spawning, or dynamic script evaluation.
Sanitization: There is no evidence of input validation or sanitization to prevent malicious instruction embedding within the encrypted or decrypted payloads.
[EXTERNAL_DOWNLOADS]: The skill requires an external Node.js library to function.
Evidence: References the @boostbrothers/react-native-aws-kms package.
Context: The package is provided by the recognized vendor (boostbrothers) and is treated as a trusted dependency for the skill's intended purpose.
[DATA_EXFILTRATION]: The skill performs network operations to reach AWS infrastructure.
Evidence: The init function configures communication with AWS KMS endpoints based on the provided region.
Status: This is part of the core functionality; no unauthorized data transfer to non-whitelisted or suspicious domains was identified.
[CREDENTIALS_UNSAFE]: The skill's initialization method handles sensitive AWS access keys.
Evidence: The API includes parameters for accessKey, secretKey, and sessionToken.
Mitigation: The documentation provides a clear security warning against hardcoding secrets in source code and recommends using dynamic identity providers or STS tokens instead.

react-native-kms-module