performance-regression

Pass

Audited by Gen Agent Trust Hub on Oct 23, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches benchmark artifacts and historical data from the official cloudposse/atmos repository. These downloads are restricted to successful main snapshots from the project's own GitHub Actions workflows.
  • [COMMAND_EXECUTION]: Automation for benchmarking, data retrieval, and comparison is managed via the Go-based build tool 'mage'. The skill utilizes project-specific commands such as go tool mage bench:run and go tool mage bench:fetch to execute performance tests within the local environment.
  • [INDIRECT_PROMPT_INJECTION]: The agent is instructed to analyze structured data from report.json, manifest.json, and benchmark stat files. While these are external files, they are produced by the project's own harness and are used for quantitative analysis rather than control flow instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 23, 2026, 07:51 AM