Content
71%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A compact, well-structured RAG assistant prompt with a clear answer workflow and explicit fallback handling. The main weakness is redundancy between the opening tool list, the Tools section, and the Guardrails.
Suggestions
Deduplicate the tool descriptions — present them once, either in the opening list or in the 'Tools' section, not both.
Collapse the 'Guardrails' into the numbered workflow where they overlap (e.g., cite-only and say-so rules) to remove restated instructions.
Add a brief example of the expected citation format to make the 'Cite which document(s)' step fully executable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient but includes redundancy — the three tools are described in the opening bullet list and again in the 'Tools' section, and the 'Guardrails' restate steps already in 'How to answer questions' — so it could be tightened. | 3 / 5 |
Actionability | It names specific tools and gives a concrete numbered process ('Search… Read… Compose using ONLY… Cite… If not enough, say so'), with only minor gaps such as no example citation format. | 4 / 5 |
Workflow Clarity | The 5-step 'How to answer questions' sequence is clear and includes an explicit insufficient-information checkpoint ('If the documents don't contain enough information, say so honestly'), though error-recovery guidance is light. | 4 / 5 |
Progressive Disclosure | The body is under 50 lines with no external references needed, and is organized into clear, well-labeled sections (How to answer, Skills, Tools, Guardrails), meeting the simple-skill exception for a top score. | 5 / 5 |
Total | 16 / 20 Passed |