The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The instructions direct the user to install a global Node.js package @yun918/codekb from the public npm registry. This package and its scope are not associated with a recognized trusted organization or the skill author's known infrastructure, posing a supply chain risk.
[COMMAND_EXECUTION]: The skill requires running shell commands and global CLI tools (codekb init, codekb status, codekb mcp). Since these commands are provided by an external, unverified package, they could execute arbitrary code on the host system.
[INDIRECT_PROMPT_INJECTION]: The skill initializes a semantic knowledge base by indexing local project code. This creates a surface for indirect prompt injection where malicious instructions embedded in a project's codebase (e.g., in comments or documentation) could be retrieved and executed by an AI agent using the index.
Ingestion points: Local source code files, configuration files, and documentation within the target project.
Boundary markers: No specific delimiters or safety instructions are mentioned to separate indexed data from agent instructions.
Capability inventory: The skill enables file system reads for indexing, file system writes for index storage, and network communication via the configured MCP (Model Context Protocol) server.
Sanitization: The instructions do not define any sanitization or filtering logic for the content being indexed.

codekb-init