The Agent Skills Directory

[PROMPT_INJECTION]: Susceptibility to Indirect Prompt Injection. The skill uses Chrome MCP to extract text, HTML structure, and metadata from arbitrary user-provided URLs. This content is used to generate specifications and prompts for builder sub-agents. Maliciously crafted content on the target website could potentially influence the behavior of the agent or its sub-agents.
Ingestion points: Extraction of textContent and DOM data from external websites via Chrome MCP.
Boundary markers: Absent. No delimiters or warnings are used to isolate untrusted web content in prompts.
Capability inventory: Shell command execution (npm/npx), file system writes, and dispatching parallel builder agents.
Sanitization: Absent. The skill extracts and uses content verbatim from target sites.
[COMMAND_EXECUTION]: Executes local build and type-checking commands including npm run build and npx tsc --noEmit. These operations occur within a project environment where the code and configuration are dynamically generated from untrusted web data.
[EXTERNAL_DOWNLOADS]: Automates the retrieval of binary assets (images, videos, fonts) from arbitrary remote URLs using a generated Node.js script (scripts/download-assets.mjs). This involves network requests to untrusted third-party domains.

clone-website