The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied Korean text, which creates an inherent surface for indirect prompt injection if the input text contains instructions meant to hijack the agent's logic.
Ingestion points: The skill reads user input as described in SKILL.md Section 1, Step 1.
Boundary markers: The skill provides strict 'Iron rules' in SKILL.md Section 2, requiring facts, proper nouns, and quotes to remain byte-for-byte identical to the source, which acts as a logical barrier against malicious manipulation of the output.
Capability inventory: The agent performs text transformation and generates examples; it does not include file-system write or network capabilities within the skill definition itself.
Sanitization: The skill relies on the agent's internal safety guardrails and the fidelity constraints defined in the instructions.
[EXTERNAL_DOWNLOADS]: SKILL.md Section 5 references a utility skill skills/korean-spell-check that utilizes a 'public checker'. This indicates a dependency on an external third-party service for grammar and spell checking, though no specific URL or execution command is provided in these files.

humanize-korean