jbvc/prompt-engineering
Expert guide on prompt engineering patterns, best practices, and optimization techniques. Use when user wants to improve prompts, learn prompting strategies, or debug agent behavior.
61
61%
Does it follow best practices?
Impact
Pending
No eval scenarios have been run
Passed
No known issues
Quality
Discovery
67%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
The description is competent with a clear 'Use when' clause and identifiable domain, but lacks the concrete specificity that would make it truly effective for skill selection. The actions described are high-level ('improve prompts', 'learn prompting strategies') rather than enumerating specific capabilities, and the trigger terms could be expanded to cover more natural user phrasings.
Suggestions
Add more concrete actions such as 'write system prompts, structure few-shot examples, apply chain-of-thought techniques, optimize token usage, design tool-use prompts'.
Expand trigger terms in the 'Use when' clause to include natural variations like 'system prompt', 'few-shot', 'chain of thought', 'LLM instructions', 'prompt template', or 'Claude API prompting'.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Names the domain (prompt engineering) and mentions some actions ('improve prompts', 'learn prompting strategies', 'debug agent behavior'), but these are fairly high-level rather than listing multiple concrete specific actions like 'write system prompts, structure few-shot examples, optimize token usage'. | 2 / 3 |
Completeness | Clearly answers both 'what' (expert guide on prompt engineering patterns, best practices, and optimization techniques) and 'when' (explicit 'Use when' clause covering improving prompts, learning strategies, or debugging agent behavior). | 3 / 3 |
Trigger Term Quality | Includes some relevant keywords like 'prompt engineering', 'prompts', 'prompting strategies', 'agent behavior', but misses common natural variations users might say such as 'system prompt', 'few-shot', 'chain of thought', 'LLM', 'Claude prompts', 'prompt template', or 'prompt optimization'. | 2 / 3 |
Distinctiveness Conflict Risk | The prompt engineering niche is reasonably distinct, but 'debug agent behavior' could overlap with general debugging or agent-building skills, and 'best practices' is generic enough to potentially conflict with coding or documentation skills. | 2 / 3 |
Total | 9 / 12 Passed |
Implementation
35%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill reads more like a tutorial or reference guide about prompt engineering concepts that Claude already deeply understands, rather than actionable instructions for a specific task. The content is verbose with explanations of well-known concepts (few-shot learning, chain-of-thought) and generic best practices. While examples are provided, they serve more as illustrations of concepts than as executable templates Claude would need to follow.
Suggestions
Remove explanations of concepts Claude already knows (what few-shot learning is, what chain-of-thought is) and focus only on project-specific patterns, preferred formats, or non-obvious techniques.
Add a concrete workflow with validation steps: e.g., 'When user asks to improve a prompt: 1. Identify the failure mode 2. Apply the matching pattern 3. Test with the user's example input 4. Compare outputs before recommending.'
Cut the 'Best Practices' and 'Common Pitfalls' sections entirely—these are generic knowledge Claude already has—and replace with specific, actionable decision trees or lookup tables.
Consider splitting detailed examples into a separate EXAMPLES.md file and keeping SKILL.md as a concise overview with clear references.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Extensively explains concepts Claude already knows well—prompt engineering patterns, few-shot learning, chain-of-thought, etc. are core LLM knowledge. Phrases like 'Improves accuracy on analytical tasks by 30-50%' and explanations of what few-shot learning is are unnecessary. The 'Best Practices' and 'Common Pitfalls' sections are generic advice Claude inherently understands. | 1 / 3 |
Actionability | Examples are provided and some are concrete (the Python template, the few-shot extraction example), but much of the content is descriptive rather than instructive. The skill tells Claude about prompt engineering concepts rather than giving it specific, executable procedures to follow when a user asks for help. The optimization example shows version progression but isn't copy-paste actionable. | 2 / 3 |
Workflow Clarity | The Progressive Disclosure section provides a clear sequence of escalating complexity, and the Instruction Hierarchy gives a useful ordering. However, there are no validation checkpoints—no guidance on how to verify a prompt is working, no feedback loops for iterating when results are poor, and no concrete steps for the prompt optimization process beyond 'test and iterate.' | 2 / 3 |
Progressive Disclosure | Content is organized into sections with headers, but everything is inline in one large file with no references to external resources. For a skill this long (~150+ lines), some content like the detailed examples or the template systems section could be split into separate files. The structure is reasonable but the monolithic approach hurts navigability. | 2 / 3 |
Total | 7 / 12 Passed |
Validation
100%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 11 / 11 Passed
Validation for skill structure
No warnings or errors.
Reviewed
Table of Contents