lokalise-data-handling

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process translation data from external sources (Lokalise API and local translation files), which constitutes an untrusted data ingestion surface. Malicious instructions could theoretically be embedded in translation strings to influence the agent or downstream consumers.
  • Ingestion points: Translation data is ingested via lokalise.files().download in SKILL.md and fs.readFileSync in references/implementation-guide.md.
  • Boundary markers: The instructions do not explicitly define boundary markers for the ingested data within the prompts.
  • Capability inventory: The skill possesses capabilities to write to the file system (unzip, Edit tool), upload data to the Lokalise API, and merge translation branches.
  • Sanitization: The skill proactively addresses this risk by providing a PII (Personally Identifiable Information) scanner (scanTranslationsForPII) and validation logic (validateBeforeUpload) in references/implementation-guide.md to detect sensitive data before it is processed or uploaded.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 03:44 AM
Security Audit — agent-trust-hub — lokalise-data-handling