Follow-ups
AuraScore 79/100

Multimodal Asset Calibration Follow-Up Protocol

Synthesize visual generation review sessions into a structured technical follow-up brief for diffusion model tuning.

Use this template following technical model calibration reviews with stakeholders to document prompt syntax modifications, weighting adjustments, and output quality thresholds. It ensures engineering and creative teams align on visual generation parameters before executing secondary render batches.

Template

Role: Principal Multimodal AI Solutions Architect with 12 years of experience in computer vision, diffusion workflows, and prompt engineering.

Context

  • Client Identity: {{client_name}}
  • Model Architecture: {{base_model_architecture}}
  • Syntax Findings: {{prompt_syntax_findings}}
  • Visual Discrepancies: {{aesthetic_mismatch_notes}}
  • Milestones: {{evaluation_milestones}}
  • Production Window: {{turnaround_window}}

Task

Draft a comprehensive post-session follow-up brief that translates subjective image evaluation feedback into definitive prompt syntax rules, hyperparameter adjustments, and an actionable revision timeline for {{client_name}}.

Method

  1. Deconstruct {{aesthetic_mismatch_notes}} into prompt weight anomalies, token clipping issues, or guidance scale misconfigurations.
  2. Cross-reference identified visual artifacts against the capabilities and token limits of {{base_model_architecture}}.
  3. Formulate revised positive prompt templates incorporating the recommendations from {{prompt_syntax_findings}}.
  4. Define explicit negative prompt blocks to suppress undesired rendering artifacts and compositional drifts.
  5. Calibrate recommended generation parameters including CFG scale, sampling steps, and denoising strength ranges.
  6. Structure a staged testing schedule aligned directly with {{evaluation_milestones}} and {{turnaround_window}}.
  7. Outline precise acceptance criteria for the upcoming batch validation review.

Constraints

  • MUST express visual feedback as reproducible prompt tokens, weights, or numeric inference parameters.
  • MUST NOT use ambiguous creative jargon without attaching explicit diffusion parameters (e.g., CFG, LoRA alpha, seed variations).
  • Action items MUST include designated ownership and target execution timestamps.
  • Keep recommendations strictly compatible with {{base_model_architecture}}.

Output format

Provide a technical follow-up brief structured in 4 mandatory sections:

  1. Executive Summary & Calibration Objectives (max 120 words)
  2. Prompt Syntax & Parameter Matrix (table containing Parameter, Current Setting, Recommended Adjustment, Rationale)
  3. Revised Prompt Schema (Positive Template, Negative Token String, Parameter Bounds)
  4. Staged Rollout & Milestone Timeline (bulleted chronological list under {{turnaround_window}})

Self-review

  • Ensure every visual discrepancy mentioned in {{aesthetic_mismatch_notes}} has a corresponding parameter or prompt fix.
  • Confirm all 6 context variables are accurately incorporated into the brief.
  • Verify the prompt syntax conforms precisely to {{base_model_architecture}} parsing rules.
AuraScore breakdown
79/100Provisional
Instruction clarity15/15 · Strong

Explicit role, a named task, and discrete steps the model can follow.

Context architecture12/12 · Strong

Background, inputs and variables the model needs before it starts.

Constraint engineering10/12 · Adequate

Hard boundaries — what the model must and must not do.

Output specification6/14 · Thin

A named, field-level shape for the response.

Reasoning structure10/10 · Strong

Ordered work items that force analysis before an answer.

Model compatibility10/10 · Strong

Length and structure that travel across frontier models.

Token efficiency5/10 · Thin

Signal density — instruction weight without padding.

Reusability7/7 · Strong

Documented variables so the scaffold adapts to new inputs.

Robustness3/5 · Adequate

Quality bar, assumptions and behaviour when inputs are thin.

Observed performance1/5 · Thin

How much real usage the template has behind it.

emails
emails-follow-ups
image-multimodal-prompting
image-generation
prompt-engineering
diffusion-models