The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script references/scripts/ExcaliMath.md uses the new Function() constructor within the generateGraphPoints function to evaluate mathematical formulas entered by the user. This creates a risk of arbitrary code execution if the input is manipulated by an attacker (e.g., via indirect prompt injection).
[EXTERNAL_DOWNLOADS]: The script references/scripts/Auto Layout.md dynamically loads and executes the elkjs library from https://cdn.jsdelivr.net/npm/elkjs@0.8.2/lib/elk.bundled.min.js at runtime using a manually created script tag appended to the document head.
[INDIRECT_PROMPT_INJECTION]: Several scripts create a significant attack surface for indirect prompt injection:
Ingestion points: references/scripts/Excalidraw Writing Machine.md reads text from canvas elements and file sections; references/scripts/Capture Note.md extracts content from searched files and canvas elements.
Boundary markers: None are present to distinguish user content from instructions when processing canvas data.
Capability inventory: The skill can create and modify files (app.vault.create, app.vault.modify) and interact with AI APIs (references/scripts/ExcaliAI.md).
Sanitization: Minimal sanitization is applied (e.g., stripping markdown comments), which is insufficient to prevent complex logic injection attacks.
[DATA_EXPOSURE]: The script references/scripts/Invert colors.md contains a documentation link to https://excalidraw-obsidian.online/wiki/color-palette. This domain has been flagged as malicious by automated security scanners. While this is currently only a reference in a comment, it represents a risk to users who visit the link.

excalidraw-automate