The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted meeting transcripts, creating a surface for indirect prompt injection attacks.
Ingestion points: Untrusted transcript data is ingested through the transcript input in SKILL.md and interpolated into the {{text_content}} placeholder in references/prompts.md.
Boundary markers: The prompt template in references/prompts.md lacks clear delimiters (e.g., XML tags or triple quotes) or specific instructions to the model to ignore any embedded commands within the transcript text.
Capability inventory: The skill does not include any scripts, command execution, network access, or file-writing capabilities, which mitigates the risk of an injection escalating into a system compromise.
Sanitization: No input validation, filtering, or escaping is performed on the provided transcript content.
[SAFE]: The skill contains no executable code or scripts; all logic is provided through natural language instructions and prompt templates.
[SAFE]: The repository reference github.com/aipoch/medical-research-skills is a legitimate vendor resource belonging to the skill's author and shows no signs of typosquatting or malicious intent.

meeting-minutes-generator