cmux-localization

Pass

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill contains legitimate instructions for managing localization strings in a development environment. The provided search patterns and rules align with standard software engineering practices.
  • [COMMAND_EXECUTION]: The skill references the use of git diff and ripgrep (rg) to identify non-localized strings in .swift, .ts, .tsx, and .md files. These commands are used for local searching and auditing purposes and do not involve untrusted network data or sensitive system paths.
  • [INDIRECT_PROMPT_INJECTION]: The skill evaluates source code files and documentation which could technically contain adversarial content. 1. Ingestion points: Local project files accessed via git and ripgrep commands mentioned in references/audit-workflow.md. 2. Boundary markers: Absent; the agent is instructed to parse and compare message keys directly. 3. Capability inventory: Shell command execution for repository searching and file diffing. 4. Sanitization: None; the agent relies on regex patterns to identify UI strings.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 12, 2026, 02:01 AM
Security Audit — agent-trust-hub — cmux-localization