TrinityMetrics AI and tooling post evaluation

TrinityMetrics AI and Tooling Post Evaluation Skill

Purpose

Evaluate whether a draft post meets TrinityMetrics standards for publication, with emphasis on practical usefulness, scientific integrity, and fit for regulated analytical workflows.

When to use

Use this skill when:

• reviewing draft posts about AI-assisted analysis, statistical tooling, pharmacometrics, reproducibility, or workflow design
• deciding whether a post is ready to publish, needs revision, or should not be published yet
• comparing multiple drafts against a consistent editorial standard

Do not use this skill for:

• grammar-only copyediting
• formatting-only review
• SEO-only review without assessing substance

TrinityMetrics framing

A strong TrinityMetrics post should usually do at least one of the following:

• reduce friction in a real scientific or analytical workflow
• make reasoning or decision boundaries more explicit
• improve reproducibility or auditability
• clarify how AI can assist without obscuring human responsibility
• provide a concrete, transferable pattern for regulated or high-consequence work

Evaluation dimensions

Score each dimension from 0 to 5.

1. Friction Reduction (0-5)

Does the post provide a tool, method, template, or pattern that meaningfully reduces real workflow friction?

Consider:

• Is there a real bottleneck being addressed?
• Would this save time, reduce confusion, or lower process overhead?
• Is the gain practical rather than merely conceptual?

2. Usability (0-5)

Can a reader apply the idea within about 10 minutes?

Consider:

• Is the setup clear?
• Is any code readable and well-scoped?
• Are inputs and outputs clear?
• Could a technically competent reader try this without guesswork?

3. Concreteness (0-5)

Is the problem and solution specific and grounded in a real use case?

Consider:

• Is the post about an actual workflow, artifact, or decision point?
• Are claims tied to examples?
• Does it avoid vague AI-generalities?

4. Scientific Integrity (0-5)

Does the post preserve sound scientific reasoning?

Consider:

• Are limitations stated clearly?
• Are risks, failure modes, or misuse scenarios identified?
• Is analysis separated from decision-making?
• Does the post avoid overstating what AI or tooling can justify?

5. Reproducibility (0-5)

Could a reader reproduce the example or pattern?

Consider:

• Are the steps clear?
• Are assumptions explicit?
• Are examples sufficiently specified?
• Would another analyst be able to repeat the workflow?

6. Signal-to-Noise (0-5)

Is the post concise and focused on the tool, pattern, or argument?

Consider:

• Does each section earn its place?
• Is there excessive framing, repetition, or abstraction?
• Is the main takeaway obvious by the end?

Optional fit checks

In addition to scoring, note whether the post:

• fits the TrinityMetrics audience of pharmacometricians, biostatisticians, statistical programmers, or clinical data scientists
• explains why the topic matters in regulated or high-accountability environments
• distinguishes useful AI assistance from automated judgment

Scoring interpretation

• 25-30: Publish
• 18-24: Revise
• Below 18: Do not publish

A high score should reflect not just polish, but usefulness and integrity.

Output format

Return the review in this structure:

• Total Score: X / 30
• Decision: Publish | Revise | Do not publish
• One-sentence verdict: Brief summary of whether the post is worth publishing and why
• Dimension Scores:
  • Friction Reduction: X / 5
  • Usability: X / 5
  • Concreteness: X / 5
  • Scientific Integrity: X / 5
  • Reproducibility: X / 5
  • Signal-to-Noise: X / 5
• What the post does well:
  • Bullet list
• Key Issues:
  • Bullet list
• Suggested Fixes:
  • Bullet list ordered from highest to lowest impact
• Best excerpt or idea to preserve:
  • Quote or short paraphrase

Review guidance

Prioritize concrete, high-impact feedback.

When possible:

• point to specific passages
• separate structural problems from minor wording edits
• identify unsupported claims or missing caveats
• recommend the smallest set of changes needed to make the post publishable

Default to substance over style.

Categorization requirement

• Classify posts as principle and/or tool.
• Classify as ai if about ai.
• Classify as biostats or pmx when either are referenced.
• Keep a tally of how many "tools" and "principles" there are. Aim for 3-4x more tools than principle posts and point this out if there are too many "principle" posts.