The Agent Skills Directory

[SAFE]: The skill's instructions and function call chain are consistent with its described purpose of document summarization and visual mapping. No suspicious external dependencies or obfuscated code were detected.\n- [PROMPT_INJECTION]: The skill ingests untrusted text from documents, creating a surface for indirect prompt injection. This is an expected risk factor for summarization tools and is managed through structured prompting.\n
Ingestion points: Document content is read using read_file as part of the summarizeToMermaidCommand in main.ts.\n
Boundary markers: The prompt provided in promptUtils.ts uses instructional headers to define the AI's role and rules, which helps separate operational logic from the processed document content.\n
Capability inventory: The skill utilizes read_file, write_file, and mkdir_p to handle document input and diagram storage, which are appropriate for its functionality.\n
Sanitization: The workflow includes automated syntax validation and repair for the generated Mermaid code using refineMermaidBlocks and checkMermaidErrors.

notemdpro-mermaid-summarizer