feat: add human-text-discipline skill for human-read text

[keenvc]

What

Adds human-text-discipline, an agent-only skill that lists the AI tells to cut from text a person reads: pull request bodies, commit messages, and captain-facing messages.

It also names the other failure most such lists miss, because flat voiceless prose is as obviously machine-made as purple prose.

Why

agent-doc-discipline covers documents an agent consumes to act.

Nothing covered the text the captain actually spends his attention on.

Every tell here is checkable: it names the observable pattern and the fix, so a reviewer can decide pass or fail from the text rather than from taste.

Scope

• AGENTS.md section 13 gains the load trigger, and section 9 gains a one-line pointer at the point where captain-facing text is written.
• Section 9 stays the owner of what those messages must contain and how internal terms are translated; this skill owns only how the prose reads, and does not restate it.
• The skill is classified agent-runtime in docs/documentation-audiences.json.
• No code changes.