P2: Glossary + vocabulary lock + diagram color legend

[ramdhanyk]

Source: 2026-04-20 multi-expert AADM audit. Docs-UX auditor.

Problem

Same concept, multiple names across the docs:

• "session" / "task" / "loop" / "round"
• "skill" / "command" / "slash command"
• "reconcile" / "coverage check" / "traceability gate" / "merge gate"
• "sprint-close" / "structural gate" / "anti-skip gate" / "lock"
• "failures log" / "prevention rules" / "institutional memory loop"

Three diagrams (START-HERE, README, handbook) use red with opposite valences (red = "good" in handbook, "refusal" in README).

Fix

• Ship GLOSSARY.md (one paragraph per term: stable ID, Satisfies chain, [DEFERRED], failures log, sprint-close, .lock, autonomy levels, test matrix A–E).
• Pick canonical terms; grep-and-replace across all docs.
• Lock a 4-color diagram legend (gate / refusal / ceremony / human-escape-hatch); apply to all three diagrams; put the legend once in START-HERE and reference from each diagram.
• Optional: tooling/scripts/docs_lint.py flagging forbidden synonyms.

Acceptance

A make docs-lint (or equivalent) flags forbidden synonyms. All three diagrams use the same legend.