R022 · llm-clarity
LLM: instructions unambiguous
- Code: R022
- Slug:
llm-clarity - Tier: llm
- Default severity: warning
- Packages: recommended, strict
- Locked: No
- Doc: https://doraval.dev/reference/rules/R022
| What | LLM-tier: skill instructions should be unambiguous — one clear meaning per rule, no foggy or self-conflicting guidance. |
| Why | Essay-like skills hide ambiguity. Agents cannot reliably “read between the lines” across many invocations. |
How to fix
- Split multi-topic skills or move side topics into
references/. - Replace hedging with a decision or condition (R019).
- If two paths are valid, state the decision rule that picks between them.