The Agent Skills Directory

[SAFE]: The skill provides comprehensive and secure implementation patterns for real-time web streaming. The documentation includes a 'Red Flags' section that correctly identifies and warns against common security risks, including credential exposure in query strings and resource exhaustion from unclosed connections.\n- [INDIRECT_PROMPT_INJECTION]: The skill outlines how to ingest and process data from external server-sent event streams. While this capability establishes a data ingestion surface, the instructions include specific mitigations to ensure the integrity of the processed data.\n
Ingestion points: Data ingestion occurs through EventSource.onmessage and fetch stream readers as documented in examples/core.md and examples/fetch-streaming.md.\n
Boundary markers: The provided parser logic correctly handles the SSE wire format using double newline (\\n\\n) message delimiters.\n
Capability inventory: The skill utilizes standard browser and network APIs for streaming but does not include capabilities for arbitrary command execution, privilege escalation, or sensitive file system access.\n
Sanitization: The skill recommends defensive programming practices, including wrapping JSON.parse in try/catch blocks and performing runtime validation of message content before use.

web-realtime-sse