translation-doc

Layer 2-5 owner for markdown / technical documentation in translation-toolkit. Reads an intake-spec from translation-intake (Layer 1), parses the markdown AST so structural spans (code, URLs, HTML, math, frontmatter, mermaid, ASCII diagrams) are masked, runs the shared DRAFT / REFLECT / IMPROVE core loop with cross-chunk <TRANSLATE_THIS> windowing, and reassembles a roundtrip-clean target file.

Inputs

• source — markdown file path (.md / .markdown / .mdx) or an inline markdown string.
• intake-spec — from translation-intake. If invoked directly without one, run intake first (--intake flag) before continuing.
• glossary_path (optional) — repo-local project glossary (L1 tier of the 4-tier resolver). Defaults to <caller_repo>/docs/i18n/glossary-{target_locale}.md; missing → fall through to L2 bundled glossary in glossary/.

Service-interface contract (the four shared fields plus web_search) is defined in the design spec §Service Interface; no per-skill duplication here.

When to use

• Routed in by using-translation-toolkit when the user supplies a .md / .mdx / .markdown file or pastes technical-documentation prose.
• Invoked directly by name when the user already knows markdown is the right specialist.
• Not for i18n format files (use translation-i18n), ad copy / catchphrases (use translation-creative), or reviewing an existing translation pair (use translation-audit).

Web search policy