oss-review
Pass
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill accesses configuration files and project manifests within the
~/.claude/plugins/config/claude-for-legal/directory. These operations are scoped to the intended functionality of the legal review plugin and do not involve access to sensitive system directories like SSH keys or environment secrets. - [SAFE]: The skill is designed to process untrusted data from external sources, including project dependency lists and the text of LICENSE or README files. While this represents a theoretical surface for indirect prompt injection, the skill implements a strict, multi-step workflow with mandatory human/attorney review checkpoints for high-risk findings, which effectively mitigates the impact of potential malicious instructions embedded in project files.
- [SAFE]: The instructions include a "No silent supplement" rule and a citation tagging system (e.g.,
[OSI],[model knowledge — verify]). These measures enforce transparency and source verification, preventing the agent from silently introducing unverified information or executing unauthorized web searches.
Audit Metadata