The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted learning materials (Markdown, plain text, or notes) which could contain embedded instructions to manipulate the agent during the extraction phase.
Ingestion points: Materials are read via scripts/generate_flashcards.py using user-provided paths or standard input.
Boundary markers: The agent workflow defined in SKILL.md does not specify the use of delimiters or 'ignore embedded instructions' warnings for the material extraction process.
Capability inventory: The skill uses the Python standard library for text parsing and CSV generation; it does not perform network exfiltration, access sensitive system files, or execute arbitrary shell commands from external data.
Sanitization: The script performs basic conversion of Markdown formatting into HTML for the output CSV but does not validate or filter the input text for malicious prompt instructions.

flashcard-studio