The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user-provided data (Korean lyrics) to perform phonological analysis and syllable counting. This creates an attack surface for indirect prompt injection where malicious instructions could be embedded within the lyrics.
Ingestion points: User-provided lyrics and linguistic data enter the agent's context through fields such as lyrics[].lines and prosody.sections[] as specified in the LYR-SPEC documentation.
Boundary markers: The skill encourages the use of structured fields (e.g., prosody.sections[].syllables, rhyme.che), which offers some structural isolation, though it does not provide explicit instructions to the agent to disregard natural language commands found within the lyrics themselves.
Capability inventory: The skill files consist of documentation and instructions; they do not contain tool definitions, shell commands, or network operations. However, the resulting analysis is intended to guide an agent which may have access to other capabilities.
Sanitization: There are no explicit instructions or scripts provided to sanitize or validate the input lyrics for hidden instructions or malicious content.

lw-korean