sanity-bench
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns detected. The skill correctly utilizes vendor-specific tools and repositories (sanity-io/sanity) for its intended purpose of benchmarking. All documented commands (pnpm, gh, node) are standard for the development environment and target trusted infrastructure.
- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing untrusted data. Evidence Chain: 1. Ingestion points: PR labels (trigger:perf-bench) and workflow dispatch inputs (ab_from, ab_to, backfill_sha). 2. Boundary markers: Instructions provided for specific maintainer-triggered labels. 3. Capability inventory: Execution of local build scripts (pnpm) and remote workflow dispatch (gh). 4. Sanitization: The skill notes that the CI workflow validates inputs, such as requiring full 40-character SHAs.
Audit Metadata