The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill maintains a persistent learning history by reading user-provided exercise answers, self-reports, and notes. Malicious instructions embedded in these historical records could influence the agent's behavior in future sessions when the history is retrieved.
Ingestion points: The script scripts/learning-store.mjs retrieves stored JSON data from the local data store (located at ~/.local/share/language-learning/ by default) during initialization and context retrieval.
Boundary markers: The instructions do not define clear delimiters or use "ignore embedded instructions" warnings when the agent processes the content of retrieved historical events.
Capability inventory: The skill utilizes scripts/learning-store.mjs for file system read/write operations and refers to the use of document rendering tools for creating workbooks.
Sanitization: While the data management script performs structural JSON validation and schema checks (ensuring IDs are valid slugs and types are expected), it does not perform linguistic sanitization or filtering of the text content to prevent embedded prompt injection.
[COMMAND_EXECUTION]: The skill relies on a bundled Node.js utility (scripts/learning-store.mjs) to handle its data persistence. The script is executed via the command line to initialize the environment, retrieve context, and record learning events. It implements security best practices such as restrictive file permissions (0o600 for files, 0o700 for directories) and strict slug-based validation for identifiers to mitigate path traversal risks.
[EXTERNAL_DOWNLOADS]: The skill provides links to external learning resources from the Japan Foundation's official Irodori website. These references are used for pedagogical purposes, such as sourcing authentic kana and handwriting examples.

japanese-learning