The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill is designed to search for, recommend, and execute external code from the npm registry. It specifically generates configurations that use npx to run these packages. This introduces a supply-chain risk where the agent might inadvertently execute malicious or typosquatted third-party code.
[DATA_EXFILTRATION]: The workflow requires reading from sensitive local directories, including ~/.claude-marketing/credentials/ and ~/.claude-marketing/brands/. These paths are documented as containing brand profiles and agency credential sets, which represent high-value targets for data exposure.
[COMMAND_EXECUTION]: The skill performs automated connectivity testing by starting MCP servers and executing tool operations. This involves running shell commands and performing network-connected operations based on dynamically discovered package configurations.
[CREDENTIALS_UNSAFE]: The skill explicitly accesses dedicated credential management directories (~/.claude-marketing/credentials/) to load and map authentication profiles for different brands.
[EXTERNAL_DOWNLOADS]: The skill queries the npm registry and other external MCP directories to evaluate and retrieve package metadata and configuration details at runtime.

add-integration