Generate next-turn prompts from winner code and evaluation results.
npx skills add https://github.com/MinhOmega/marlin-skill --skill checkpoint-prompt
Please help me install this Agent Skill. Skill: checkpoint-prompt Source: https://github.com/MinhOmega/marlin-skill/tree/main/.claude/skills/checkpoint-prompt Command: npx skills add https://github.com/MinhOmega/marlin-skill --skill checkpoint-prompt
The Checkpoint Prompt skill provides a structured mechanism to generate the next-turn prompt that addresses gaps identified in the current evaluation, ensuring improvements are concrete, testable, and compliant with the promotion of quality across agent prompts.
Run the next-turn prompt generation against the active evaluation workspace and save the result as turn_N/turn_N_next_prompt.md
High-intent search queries and answers about installing and using this skill.
To generate a next-turn prompt, the skill analyzes accepted_baseline.json and turn evaluation results to produce flowing prose targeting concrete file paths and functions, saving the output to turn_N_next_prompt.md.
Automating prompt generation for iterative code improvements involves reading prior prompts and evaluation results to craft targeted, testable instructions that enforce guardrails and address specific implementation gaps in the winner's code.
Yes, checkpoint prompts ensure code changes remain localized by generating instructions with explicit references to relevant code locations, file paths, and tests, enforcing guardrails for specific and testable modifications.
Yes, you need prior evaluation results. The skill reads accepted_baseline.json, turn_N_evaluation.md, and all prior prompts to analyze gaps and produce a flowing prose prompt targeting concrete code locations.
Use automated next-turn prompt generation when an evaluation reveals implementation gaps and you need to systematically guide the next turn toward new tests, edge cases, and concrete changes without manual prompt crafting.
understudylabsGenerate harness, metric, splits, and baseline artifacts from workloads.
kkestellExecute repository implementation plans with code changes, tests, and reviews.
LinusConradMGenerate a Quarto report from the latest LLM evaluation results JSON.
zeveckAutomate end-to-end verification of code changes with diff review, test execution, and UI checks.
