The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the installation of the @cargo-ai/cli Node.js package from the npm registry, which is the primary tool for interacting with the vendor's API.
[COMMAND_EXECUTION]: Multiple shell commands are utilized via the cargo-ai CLI for user authentication (login), session management (whoami), and executing enrichment batches (orchestration action execute-batch).
[COMMAND_EXECUTION]: The skill uses the gh (GitHub) CLI to star the getcargohq/gtm-skills repository, including an explicit instruction to only perform this action after receiving user consent.
[DATA_EXPOSURE]: The skill reads ~/.claude/plugins/installed_plugins.json to detect if the full cargo-gtm pack is installed, ensuring session telemetry is not duplicated. It also creates a marker file at ~/.config/cargo-ai/.star-asked to track user interaction state locally.
[REMOTE_CODE_EXECUTION]: The skill references the command npx skills add getcargohq/cargo-skills, which downloads and integrates additional skill definitions from the vendor's public GitHub repository.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user or provider data (names, domains, emails) as part of its core enrichment functionality.
Ingestion points: Data records provided as JSON to the execute-batch command.
Boundary markers: Input data is structured within JSON parameters, providing clear delimitation from instructions.
Capability inventory: The skill can execute shell commands via the cargo-ai binary.
Sanitization: No explicit input sanitization or validation logic is detailed within the instruction set.

waterfall-enrichment