Markdown Doc Multilingual Translator

What to check before you install or run this skill: 1) Credentials & privacy: The SKILL.md/README expect a Claude/Anthropic API key (via --api-key or env var) and network access for translation. The registry metadata did not declare required env vars — inspect translator.py to confirm exact env var names (e.g., CLAUDE_API_KEY, ANTHROPIC_API_KEY) and endpoints. If you will translate sensitive or proprietary docs, prefer a local/offline translation mode or review terms of the external API provider. 2) Dependencies: The package expects Python dependencies (requests, pyyaml, markdown, and an LLM client). Ensure those libraries come from trusted sources in your environment and are pinned if you install them yourself. 3) Inspect network calls: Open scripts/translator.py and any code that does HTTP requests to confirm what is sent to external servers (full document text vs. segmented or redacted snippets), whether logs include original content, and what endpoints are used. 4) Test with non-sensitive data: Run translations on safe sample files first to confirm behavior (where requests go, what is logged, and if URLs or code blocks are preserved). Use --validate and --verbose to inspect reports. 5) Glossary & data files: Review included glossaries and any references for private data; custom glossaries are loaded from disk and may be merged — ensure they don't inadvertently include secrets. 6) Ask the publisher: Because the package source/homepage is unknown and the registry metadata omits expected env vars, prefer to use only if you can audit the code or get a trustworthy source. If you cannot audit, treat networked translation as a potential data-exfiltration risk and avoid processing sensitive documents.