The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests component data and source code from the external 21st.dev registry, which could potentially contain malicious instructions intended to influence the behavior of the AI agent.
Ingestion Points: Remote data is retrieved via mcp__21st-dev-magic__search, mcp__21st-dev-magic__get_component, and fallback WebFetch operations targeting 21st.dev (SKILL.md).
Boundary Markers: The instructions do not define explicit delimiters or security warnings to the agent to treat external component code as untrusted data or to ignore embedded instructions.
Capability Inventory: The skill uses Read, Glob, and Grep to inspect local project files and WebFetch/WebSearch for network access, providing a mechanism where ingested instructions could potentially trigger agent tool usage (SKILL.md).
Sanitization: No evidence of sanitization or safety filtering is applied to the fetched component source code or metadata before it is presented to the agent.
[EXTERNAL_DOWNLOADS]: The skill is designed to fetch external React components and related metadata from the 21st.dev registry. While this is its primary function, it involves downloading third-party content into the project context.
[COMMAND_EXECUTION]: The skill retrieves and displays installCommand strings sourced from a remote registry. This creates a vector where malicious registry content could provide dangerous commands for the user or agent to consider.

component-search