sanity-radar-investigate

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources such as Studio Radar prompts and benchmark PR comments to trigger investigation procedures.
  • Ingestion points: Triggers include "investigation prompts" from Studio Radar, drift flags, and performance benchmark comments on GitHub PRs.
  • Boundary markers: The skill does not specify explicit delimiters to separate untrusted data from instructions.
  • Capability inventory: The skill uses the GitHub CLI (gh workflow run, gh run watch) to trigger remote actions and pnpm (build:bench, bench run) to execute local benchmarking scripts.
  • Sanitization: The procedure guides the agent to extract specific structured data like commit SHAs and scenarios, which provides structural validation, although raw strings from external prompts are processed.
  • [COMMAND_EXECUTION]: The skill utilizes the GitHub CLI and pnpm to automate the investigation process, including triggering workflows and running local benchmarking sessions. These are standard developer operations for the sanity-io/sanity repository.
  • [EXTERNAL_DOWNLOADS]: The skill interacts with the official sanity-io/sanity GitHub repository to fetch data and compare code versions. These interactions are consistent with the vendor's own infrastructure and established development practices.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 06:48 AM
Security Audit — agent-trust-hub — sanity-radar-investigate