translating-ios
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for processing external project files, including String Catalogs (
.xcstrings), source code, and build settings. This creates an inherent attack surface where malicious instructions could be embedded in localization data or developer comments (Indirect Prompt Injection). However, the skill mitigates this by instructing the agent to focus on semantic fidelity, structural validation, and explicit QA gates. - Ingestion points: Instructions involve reading and analyzing String Catalogs and application source code as mentioned in
SKILL.md. - Boundary markers: While explicit delimiters for untrusted data are not mentioned, the skill requires reconciling catalog entries with actual call sites and project-specific release policies.
- Capability inventory: The skill describes capabilities for compiling catalogs, extracting targets, and performing runtime checks on localized strings (mentioned in the 'QA Handoff' section).
- Sanitization: The instructions explicitly require rejecting raw keys, runtime concatenation, and invalid locale formatting to prevent unintended output behavior.
Audit Metadata