The Agent Skills Directory

[PERSISTENCE]: The skill uses scripts/profile_store.py to maintain state across sessions by writing to local configuration files in the user's home directory (e.g., ~/.config/agent-skills/profile.json). This is used to store 'durable decisions' and preferences.
[DATA_EXFILTRATION]: The skill accesses local configuration files to retrieve user identity, brand preferences, and workspace settings. While it implements a filter to prevent the storage of sensitive keys such as 'token' or 'password', it represents a mechanism for the agent to access and store user-specific information.
[DYNAMIC_EXECUTION]: The script scripts/audit_html_fonts.py utilizes the Playwright library to launch a Chromium browser instance for introspecting font rendering in HTML/CSS templates via the Chrome DevTools Protocol.
[EXTERNAL_DOWNLOADS]: The illustration workflow involves downloading images and assets from external sources using curl based on entity extraction and search results derived from the input document.
[INDIRECT_PROMPT_INJECTION]: The skill possess a surface for indirect prompt injection as it ingests untrusted Markdown documents and uses their content to generate prompts for image generation. 1. Ingestion points: Markdown files provided as the file_path argument. 2. Boundary markers: The skill lacks explicit delimiters or instructions to ignore embedded commands within the extracted entities. 3. Capability inventory: Network access (curl), browser automation (playwright), and file writing to the attachments/ directory. 4. Sanitization: There is no explicit sanitization or validation of the extracted document content before its interpolation into AI generation prompts.

lov-illustrate