The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The login tool accepts raw username and password as input parameters. These are used to authenticate the user on the iconfont.cn website through browser automation. The implementation stores these credentials only in volatile memory during the login process and does not persist them to disk.
[COMMAND_EXECUTION]: The skill executes commands through browser automation (Playwright, Puppeteer, or MCP) to navigate pages and interact with DOM elements. It also performs file system operations, including directory creation (fs.mkdir) and file writing (fs.writeFile) to save downloaded icons.
[EXTERNAL_DOWNLOADS]: The skill defines playwright and puppeteer as optional dependencies in its package.json. These are well-known libraries required for the skill's primary browser automation functionality.
[DATA_EXFILTRATION]: The skill extracts session cookies from the browser session after a successful login to facilitate subsequent authenticated API requests to iconfont.cn. These cookies are used exclusively for accessing the platform's search and download features.
[PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it processes and displays data fetched from an external website.
Ingestion points: Data is ingested from iconfont.cn search results and icon metadata via the search and fetchSVGFromDetailPage functions in scripts/index.js.
Boundary markers: There are no explicit delimiters or protective instructions used when presenting icon metadata or lists to the agent.
Capability inventory: The skill can write files to the local file system (fs.writeFile), perform network requests (fetch) using session cookies, and execute scripts within a browser context.
Sanitization: The skill includes a sanitizeFileName function to remove illegal characters from icon names used as filenames, which helps prevent basic path traversal via the filename; however, the outputPath parameter allows for arbitrary directory specification.

iconfont-downloader