The Agent Skills Directory

[COMMAND_EXECUTION]: Executes local Python scripts (connector-status.py, resolve_model.py) and shell commands to verify the integrity of brand directories and configuration files.
[EXTERNAL_DOWNLOADS]: Performs network connectivity checks via curl to URLs defined in the brand's .mcp.json configuration file to verify the reachability of external MCP servers.
[DATA_EXFILTRATION]: Accesses sensitive locations including ~/.claude-marketing/{brand}/credentials.json to verify credential storage health. This access is limited to existence checks, and the skill includes explicit instructions to avoid logging or printing actual secret values.
[PROMPT_INJECTION]: Processes external content from configuration files (brand-profile.json, .mcp.json) to populate validation reports. This creates an indirect prompt injection surface because untrusted data is interpolated into the agent's output without explicit sanitization or boundary markers.

validate-profile