latest-version
Pass
Audited by Gen Agent Trust Hub on Mar 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Node.js script (
scripts/latest.js) to perform version lookups. - [EXTERNAL_DOWNLOADS]: The script performs network requests to retrieve package metadata and documentation from well-known and trusted services.
- Fetches package information from the NPM registry (
registry.npmjs.org). - Retrieves project details from PyPI (
pypi.org). - Queries the Go proxy (
proxy.golang.org) and the GitHub API (api.github.com) for Go package information. - Fetches crate metadata from Crates.io and gem information from RubyGems.org.
- Downloads model documentation from Google's official developer site (
ai.google.dev). - [PROMPT_INJECTION]: The skill processes untrusted data from external software registries which may contain embedded instructions in package descriptions or README files (Indirect Prompt Injection surface).
- Ingestion points: External metadata is fetched from
registry.npmjs.org,pypi.org, andapi.github.comvia thescripts/latest.jsscript. - Boundary markers: External content (such as deprecation reasons or repository descriptions) is printed to the console with minimal delimiters, which could allow embedded instructions to be seen by the agent.
- Capability inventory: The skill allows the agent to retrieve data which is then used to modify local configuration files like
package.jsonorrequirements.txt. - Sanitization: No filtering or sanitization of the retrieved external strings is performed before they are displayed to the agent.
Audit Metadata