Generate improved artifact versions from ASI feedback in the simmer loop.
npx skills add https://github.com/2389-research/simmer --skill simmer-generator
Please help me install this Agent Skill. Skill: simmer-generator Source: https://github.com/2389-research/simmer/tree/main/skills/simmer-generator Command: npx skills add https://github.com/2389-research/simmer --skill simmer-generator
Generates an improved version of an artifact by applying the judge's ASI feedback from the previous round, enabling faster convergence to a high-quality candidate.
Provide the next improved artifact candidate based on the judge's ASI and current iteration context.
High-intent search queries and answers about installing and using this skill.
To improve an artifact using ASI feedback, provide the current candidate text, rubric, ASI, and iteration number to generate an updated version within a serialized refinement loop.
Yes, iterative refinement supports both single-file and workspace targets, applying the ASI-driven direction to craft an updated artifact across the entire specified workspace.
Generating an improved artifact candidate requires the current text or workspace reference, rubric, ASI, iteration number, artifact type, and optional background context to execute the refinement.
The serialized refinement loop processes ASI feedback from the previous round to generate a complete updated artifact, enabling faster convergence to a high-quality candidate in the next iteration.
No, the artifact generation executes ASI-driven direction specifically within the simmer orchestrator's iterative workflow, requiring its loop context to apply judge feedback properly.
To automate repeatable artifact improvements, run end-to-end iterations within the simmer orchestration loop, applying ASI feedback sequentially to generate refined candidates automatically.
matt-rileyDesign iterative evaluation loops with convergence guards for agent output refinement.
2389-researchScore candidate artifacts against user-defined criteria and generate an ASI.
2389-researchDispatch three judges to score, deliberate, and synthesize consensus scores with a single ASI.
1012LoninAudit novel chapters for narrative structure, character consistency, pacing, and dialogue.
