grill-skill
Pass
Audited by Gen Agent Trust Hub on Nov 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a controlled 'interview' phase before performing any filesystem modifications or reading sensitive data.
- [SAFE]: Access to sensitive session history (mining past sessions) is explicitly gated by a 'Phase 2' check that requires user confirmation before reading files from standard agent session directories (e.g.,
~/.claude/projects,~/.codex/sessions). - [SAFE]: The skill correctly identifies that login commands (e.g.,
caliper login chatgpt) involve browser-based authentication and mandates user consent before execution. - [SAFE]: Instructions for evaluating the skill itself (
grill-skill.eval.yaml) correctly useuser_customizations: falseto ensure a portable, isolated environment, which is a security best practice for evaluation workflows. - [SAFE]: The skill uses standard CLI tools (
caliper,pipx) and follows documented workflows for the platform without bypassing safety filters or using obfuscation.
Audit Metadata