prompt-optimization
Applies prompt repetition to improve accuracy for non-reasoning LLMs
28
20%
Does it follow best practices?
Impact
—
No eval scenarios have been run
Passed
No known issues
Optimize this skill with Tessl
npx tessl skill review --optimize ./agent-skills/prompt-optimization/SKILL.mdQuality
Discovery
32%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 identifies a specific technique (prompt repetition) and a target context (non-reasoning LLMs), giving it some distinctiveness. However, it lacks concrete action details, natural trigger terms users would say, and critically has no 'Use when...' clause to guide skill selection. It reads more like a brief label than a functional description.
Suggestions
Add an explicit 'Use when...' clause, e.g., 'Use when the user asks about improving accuracy for non-reasoning models, repeating prompts, or prompt engineering for models without chain-of-thought.'
List specific concrete actions the skill performs, such as 'duplicates key instructions within prompts, tests repetition strategies, and measures accuracy improvements.'
Include natural trigger terms users might say, such as 'repeat prompt', 'prompt engineering', 'boost accuracy', 'non-CoT models', or specific model names like 'GPT-4 mini', 'Claude Haiku'.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Names a specific technique ('prompt repetition') and a domain ('non-reasoning LLMs'), but describes only one vague action ('applies') without listing concrete steps or outputs like 'duplicates key instructions', 'benchmarks accuracy', etc. | 2 / 3 |
Completeness | It partially addresses 'what' (applies prompt repetition) but completely lacks a 'Use when...' clause or any explicit trigger guidance, which per the rubric should cap completeness at 2 — and since the 'what' is also quite thin, this falls to 1. | 1 / 3 |
Trigger Term Quality | Includes some relevant terms like 'prompt repetition', 'accuracy', and 'non-reasoning LLMs', but misses common user-facing variations such as 'repeat instructions', 'prompt engineering', 'improve model output', or specific model names users might reference. | 2 / 3 |
Distinctiveness Conflict Risk | The combination of 'prompt repetition' and 'non-reasoning LLMs' is fairly niche, but 'improve accuracy' is generic enough to overlap with other prompt engineering or optimization skills. | 2 / 3 |
Total | 7 / 12 Passed |
Implementation
7%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill is fundamentally problematic because it describes an automatic system that requires no action from Claude, making it essentially non-actionable. The content is padded with explanations of mechanisms, fabricated-looking performance metrics, and references to an undefined 'LOKI' system with no implementation details. It reads more like a product marketing document than an actionable skill.
Suggestions
Either provide executable code that Claude can use to implement prompt repetition itself, or remove this skill entirely if the system is truly automatic and requires no Claude involvement.
Remove the performance table and mechanism explanation—these are not actionable and waste tokens on information Claude cannot verify or use.
If this skill is meant to guide Claude's behavior, provide concrete instructions: e.g., 'When delegating to a Haiku sub-agent, repeat the core instruction twice in the prompt you construct.'
Add validation steps: how should Claude verify that prompt repetition is working or measure its impact in practice?
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill explains concepts Claude already knows (how attention works, what prompt repetition is), includes a performance table with fabricated-looking numbers that add little actionable value, and has significant padding. The 'Agent Instructions' section tells Claude things it can't act on (automatic prompt repetition), and the 'How It Works' section explains a mechanism that isn't actionable. Much of this content is filler. | 1 / 3 |
Actionability | The skill describes an automatic system ('No action needed from you') rather than providing executable guidance. The configuration section shows environment variables but doesn't explain how to actually implement the repetition. There's no code Claude can execute—just a conceptual before/after showing string concatenation and env var settings for an undefined 'LOKI' system. | 1 / 3 |
Workflow Clarity | There is no clear workflow or sequence of steps. The skill says the optimization is automatic, so there's no process for Claude to follow. There are no validation steps, no error handling, and no feedback loops. The 'When to Activate' section lists triggers but doesn't describe what Claude should actually do. | 1 / 3 |
Progressive Disclosure | The content is organized into clear sections with headers, and references a separate file (references/prompt-repetition.md). However, no bundle files are provided to verify the reference exists, and the main content includes material that could be trimmed (performance tables, mechanism explanation) rather than properly delegated to reference files. | 2 / 3 |
Total | 5 / 12 Passed |
Validation
90%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 10 / 11 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
frontmatter_unknown_keys | Unknown frontmatter key(s) found; consider removing or moving to metadata | Warning |
Total | 10 / 11 Passed | |
- Repository
- asklokesh/loki-mode
- Commit
0cf4f4d
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.