evaluate-skill

Pass

Audited by Gen Agent Trust Hub on Nov 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to run caliper commands for validating specs, executing evaluation tasks, and generating reports.
  • [EXTERNAL_DOWNLOADS]: Instructions are provided to install the caliper-eval package from a public registry using pipx. This is a necessary dependency for the skill's primary function.
  • [REMOTE_CODE_EXECUTION]: The Caliper framework can fetch and install external skills from Git repositories (such as GitHub) as part of an evaluation's environment setup.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes evaluation specifications that include prompts and criteria for other AI models. While these could contain instructions designed to influence an agent, the framework includes a sandbox feature with forbidden_files to manage risks when processing external test data.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 15, 2026, 10:26 AM