# Language Standard
This standard adapts Lore's language canon for agent-spec documentation.
## Voice
- Prefer direct, active voice.
- Use imperative mood for procedural steps.
- Use indicative mood for reference and explanation.
- Avoid unsupported claims; ground claims in code paths, KLL artifacts, specs,
public references, or the document itself.
- Avoid AI theater: do not claim a model verified behavior unless lifecycle,
tests, trace, or human review evidence exists.
## Terms
- Use `agent-spec` for the tool name.
- Use `Task Contract` for `.spec.md` task files.
- Use `KLL` only after first spelling out Knowledge Liveness Layer when the
audience may be new.
- Use `requirement replay` for evidence replay, not deterministic LLM replay.