The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends using the partial-json NPM package for robust incremental parsing of streamed LLM content.
[PROMPT_INJECTION]: The skill focuses on processing untrusted data from LLM responses. It implements defense-in-depth through regex-based sanitization (correcting trailing commas and unquoted keys) and strict schema validation to ensure the integrity of data before it is consumed by a pipeline.
[COMMAND_EXECUTION]: The provided logic performs safe string transformations and JSON parsing without invoking shell commands or executing arbitrary code.

response-parser