prompt-repetition

Prompt repetition improves LLM accuracy by re-processing prompts to reinforce context, mitigating causal-attention limitations in lightweight models. This allows a second pass over the full input, yielding more reliable results without architectural changes.

You apply prompt repetition to long-context prompts using an auto-apply mechanism with configurable repetition counts. The Skill defaults to 2x repetitions and uses a marker to ensure no duplicate application.

Yes, prompt repetition works with options-first MCQ tasks by enabling a second pass over the full input. It specifically targets these prompt-heavy scenarios to enhance response accuracy on targeted models.

The default repetition setting for position-based prompts is 3x repetitions. This configurable setting ensures better attention distribution for position-dependent queries.

Yes, you can configure repetition counts for different LLM contexts. The Skill provides configurable repetition settings with token-context management to handle various prompt-heavy tasks effectively.

You should avoid prompt repetition when token limits are a strict constraint, as it increases token usage. The Skill includes token-context management to mitigate this, but very tight context windows may still be problematic.