translation-intake

Purpose

Translates user intent into a structured 5-axis intake spec that downstream skills consume. This is the Layer 1 owner in the translation-toolkit 5-layer pipeline; subsequent layers (parse / protect / draft / reflect / improve / verify / output) are owned by the four format-specialist skills (translation-i18n / translation-doc / translation-creative / translation-audit), each of which reads the intake spec produced here.

Modes

• auto (default): single source-analysis call infers all 5 axes from source content. Zero-friction default; locale pair still user-supplied.
• explicit (--explicit / -e): user is prompted for each of 5 axes + skopos. Use when the auto inference misfired or when the user has strong opinions source content alone wouldn't reveal.

See protocols/intake-auto.md and protocols/intake-explicit.md for the per-mode pipeline / interaction details.

When to use

• Invoked by using-translation-toolkit (the router) when user input lacks clear axis signals — e.g. a short raw-text request with no format / domain hint.
• Invoked directly by the user via the --intake flag on any of the 4 active skills, when they want to inspect / lock the 5-axis spec before the format-specialist runs.
• Re-invoked with --explicit after a dissatisfying auto pass to override one or more axes interactively (auto inferences seed the prompts as defaults).