The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses natural instructional language and does not contain any patterns attempting to override agent behavior or bypass safety guidelines.
[DATA_EXPOSURE]: No hardcoded credentials, sensitive file paths, or suspicious network operations were found. The skill correctly recommends managing message IDs and timestamps within the application logic.
[OBFUSCATION]: Analysis of all files, including code snippets and documentation, revealed no signs of Base64 encoding, zero-width characters, homoglyphs, or other obfuscation techniques.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns or unverifiable dependencies were detected. The skill provides static code examples for the developer to implement using the standard PubNub SDK.
[PERSISTENCE_MECHANISMS]: The skill documents a persistent offline queue using standard browser and mobile storage APIs (IndexedDB, SQLite) for legitimate application functionality (handling network disconnects), but does not attempt to establish malicious persistence.
[INDIRECT_PROMPT_INJECTION]: The skill defines patterns for processing external message data. While it lacks explicit prompt boundary markers, the described patterns (deduplication, version routing) involve safe handling of data objects without exposing them to executable sinks or dynamic code generation. This is considered a safe application of real-time messaging patterns.

pubnub-reliability