rover
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The installation instructions for macOS, Linux, and Windows utilize piped shell execution from remote URLs (
https://rover.apollo.dev/nix/latest | shandhttps://rover.apollo.dev/win/latest' | iex). These downloads originate from the official domain of the skill author (Apollo GraphQL). - [EXTERNAL_DOWNLOADS]: The skill fetches configuration, schemas, and CLI binaries from Apollo GraphOS domains (
apollo.dev,apollographql.com). It also utilizes the official@apollo/roverpackage from the npm registry. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted GraphQL Schema Definition Language (SDL) data fetched from remote endpoints or local files.
- Ingestion points: Data enters the agent context through
rover graph fetch,rover subgraph fetch, androver subgraph introspectcommands. - Boundary markers: The skill explicitly instructs agents to use piping (e.g.,
rover graph fetch | rover schema describe -) specifically to prevent large SDL payloads from flooding the agent's context, which acts as a structural mitigation against malicious content in the schema. - Capability inventory: The agent has access to file system operations (Read, Write, Edit) and network operations (via Rover and Bash tools) to manage these schemas.
- Sanitization: The skill emphasizes using the
rover schematool to filter and search data rather than reading raw external input directly.
Audit Metadata