software-localisation

Pass

Audited by Gen Agent Trust Hub on Oct 6, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill provides architectures for ingesting and processing external translation catalogs (locales/**/*.json).
  • Ingestion points: The skill guides the agent to process localized content stored in JSON files across various frameworks.
  • Boundary markers: The documentation emphasizes the use of ICU Message Format and structured data, which helps define boundaries.
  • Capability inventory: Includes maintenance scripts scripts/check_urls.py (network access) and scripts/check_examples.py (local file read); references/ops-runbook.md provides local analysis commands using rg, jq, and node.
  • Sanitization: The skill provides explicit warnings and remediation guidance for XSS vulnerabilities associated with rendering translated content (e.g., v-html or Angular i18n pipes) in references/icu-message-format.md.
  • [EXTERNAL_DOWNLOADS]: The maintenance script scripts/check_urls.py validates the availability of documentation and specification URLs listed in data/sources.json. These URLs target well-known technology documentation sites (e.g., MDN, Vercel, Unicode, W3C) and are documented for transparency.
  • [COMMAND_EXECUTION]: The references/ops-runbook.md file contains recommended commands for local repository analysis and triage (e.g., rg, jq, wc, sed). These are standard development tools used to manage large localization catalogs.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 6, 2026, 02:03 PM