The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The setup.sh script installs several Node.js packages from the NPM registry, including @mermaid-js/mermaid-cli, beautiful-mermaid, and puppeteer. These are standard tools for the skill's purpose.- [EXTERNAL_DOWNLOADS]: The render.mjs script fetches icon datasets from unpkg.com and loads the Mermaid library from jsdelivr.net. These are well-known CDNs for web resources.- [COMMAND_EXECUTION]: The skill uses execSync to call its internal scripts and the Mermaid CLI for validation and rendering. While render.mjs uses path quoting to mitigate basic command injection, the batch.mjs script performs simpler argument joining for parallel processing.- [REMOTE_CODE_EXECUTION]: The setup.sh script downloads and installs the Chromium browser using npx puppeteer browsers install chrome to enable headless rendering. Rendering is performed by executing the mmdc CLI or a Puppeteer instance.- [INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill reads and analyzes project source code (e.g., schema files, models, and API definitions) to generate diagrams, as detailed in references/code-to-diagram.md.
Boundary markers: There are no explicit instructions or delimiters configured to prevent the agent from following Mermaid-style instructions or comments found within the user's source code.
Capability inventory: The skill has the capability to write files and execute shell commands via its rendering scripts.
Sanitization: The extraction logic relies on pattern matching and does not include specific sanitization for malicious strings that might influence the agent's behavior during analysis.- [DYNAMIC_EXECUTION]: The render.mjs script constructs a temporary HTML file containing inline JavaScript for icon pack registration and diagram rendering. This HTML is then executed within a headless Puppeteer instance.

mermaid-studio