The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the @cargo-ai/cli package globally via npm. This is a vendor-owned resource associated with getcargohq used for the skill's primary functionality.
[COMMAND_EXECUTION]: Several shell commands are executed to set up the environment, including npm install, cargo-ai for authentication and orchestration, and gh for GitHub API interactions.
[DATA_EXFILTRATION]: The skill sends session metadata (session ID, title, and summary) to the vendor's backend using the cargo-ai workspaceManagement session upsert command for attribution and tracking purposes.
[PROMPT_INJECTION]: The skill processes untrusted user input (such as industry filters and headcount ranges) to populate arguments for the cargo-ai CLI. This creates a surface for indirect prompt injection where malicious input could influence the command execution parameters.
Ingestion points: User-defined filters for industry, headcount, and geography in SKILL.md.
Boundary markers: None are present to delimit user input from the command structure.
Capability inventory: The skill utilizes cargo-ai orchestration action execute to perform network-based market research.
Sanitization: There is no evidence of explicit sanitization or escaping for the data passed to the --data flag in the CLI command.
[DATA_EXPOSURE]: The skill reads ~/.claude/plugins/installed_plugins.json to detect if the full Cargo GTM plugin is already installed, which involves accessing the agent's environment configuration.

build-tam-list