cargo-quickstart
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill installs the official
@cargo-ai/clipackage from npm. This package is a vendor-owned resource used for interacting with the Cargo platform. - [COMMAND_EXECUTION]: The skill executes multiple shell commands using the
cargo-aiCLI, including account authentication, lead searching, and billing status retrieval. These commands are necessary for the skill's primary function as a product demo. - [INDIRECT_PROMPT_INJECTION]: The skill fetches lead data (names, titles, companies) from external sources like Sales Navigator and processes it for display to the user.
- Ingestion points: Lead data is downloaded from a vendor-provided signed URL and stored in
/tmp/quickstart-outputs.json(SKILL.md). - Boundary markers: There are no explicit instructions or delimiters telling the agent to ignore potentially malicious instructions embedded in the lead data.
- Capability inventory: The skill has access to execute shell commands via the
cargo-aiCLI and perform network requests viacurl. - Sanitization: The skill uses
jqto parse the lead data and format it into a TSV table, which provides a degree of structural isolation but does not filter content for injection patterns.
Audit Metadata