score-leads
Pass
Audited by Gen Agent Trust Hub on Aug 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs a global installation of the vendor CLI using
npm install -g @cargo-ai/cli. Global installations often require elevated privileges and modify the system-wide environment. - [COMMAND_EXECUTION]: It includes a command to modify the user's GitHub account via
gh api -X PUT /user/starred/getcargohq/gtm-skillsto star the vendor's repository. While the instructions require the agent to ask for consent, this is an automated mutation of external user data. - [REMOTE_CODE_EXECUTION]: The skill utilizes
npx skills add getcargohq/cargo-skills, which downloads and executes remote code from the npm registry. - [DATA_EXPOSURE]: It inspects the local agent environment by reading
~/.claude/plugins/installed_plugins.jsonto check for specific plugin installations. This accesses internal configuration files of the agent platform. - [DATA_EXFILTRATION]: The skill transmits session metadata, including a session ID and summary, to the vendor's endpoint via
cargo-ai workspaceManagement session upsert. This is documented as being for attribution and session tracking. - [EXTERNAL_DOWNLOADS]: The skill initiates downloads of external packages (
@cargo-ai/cli) from the npm registry during its setup and update phases.
Audit Metadata