The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted markdown data from the user's local StudyVault directory to parse concepts and create quiz questions, creating an indirect prompt injection risk if the files contain hidden instructions.
Ingestion points: Phase 3 reads files matching **/StudyVault/ and concepts/{area}.md into the agent context.
Boundary markers: Absent. The instructions do not prescribe delimiters or instructions to ignore embedded prompts when reading user markdown data.
Capability inventory: Modifies local tracking files (concepts/{area}.md and dashboards) and interacts with the user via AskUserQuestion. No network or shell execution capabilities exist.
Sanitization: Absent. No escaping or content verification is performed on the ingested notes before formatting quiz questions.

tutor