The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and process text from locale dictionary files to perform translations, which introduces a potential vector for instructions hidden within the data.
Ingestion points: The agent reads engine/i18n/en.js and other locale files to identify keys and values for translation.
Boundary markers: The instructions do not provide explicit delimiters or "ignore embedded instructions" warnings for the content being translated.
Capability inventory: The skill involves file-writing operations to update locale dictionaries and the execution of local shell commands via npm scripts (npm run i18n:check, npm run i18n:report).
Sanitization: There is no mention of sanitizing or escaping the external content before the agent processes it for translation.

tinyworld-i18n