The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute shell commands, specifically using curl to download assets and npx tsc for project validation. The shell commands use dynamically extracted data from external sources, which may lead to command injection if not properly handled by the agent.
[EXTERNAL_DOWNLOADS]: External images such as logos and hero banners are downloaded from remote URLs extracted during the scraping phase and saved directly to the local file system.
[DATA_EXFILTRATION]: The skill performs outbound network requests using the WebFetch tool to retrieve HTML content from user-provided domains for theme extraction.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from external websites to drive its automated code generation loop. • Ingestion points: HTML content and CSS tokens fetched from external websites via WebFetch in Step 0a of SKILL.md. • Boundary markers: None. Scraped content is parsed and used directly to populate configuration files and component code. • Capability inventory: File system write access for application code and assets, network tool access for scraping, and shell command execution via Bash. • Sanitization: No explicit sanitization or validation steps are defined for the scraped website data. The skill generates React code using dangerouslySetInnerHTML for JSON-LD schemas and interpolates scraped tokens into globals.css, which could facilitate cross-site scripting (XSS) or CSS injection in the generated output if the source website is compromised or malicious.

build-landing-page