Agent Skill Library, kit

Substrate Builder: make AI decisions repeatable

Turn a loose classification or scoring prompt into explicit rules, then test whether different runs and models return the same answer. The download includes the skill, scripts, templates, references, and worked examples.

Who it is for

Revenue and operations teams using AI to classify leads, score accounts, route work, or make other decisions with a fixed set of answers.

Download the full kit Back to the library

The download holds all 1 SKILL.md files plus templates and setup notes (36 files).

Safari note: Safari on Mac can unzip downloads automatically. To keep the kit zipped, open Safari, go to Settings, General, and turn off "Open 'safe' files after downloading". Chrome, Edge and Firefox keep the zip as it is.

Install this package

Run the command below in your terminal. It downloads and extracts the package, then copies its skill folders into Claude Code's skills directory. Existing skill folders with the same names may be replaced.
Keep the extracted substrate-builder folder. Read its setup notes for any additional scripts, templates, or workspace steps before running the skills. For another assistant, follow its own skill installation instructions using the downloaded folders.
Start with the package's first step and check the output before using it.

curl -fL "https://www.therevenueaireport.com/agent-skills/kits/substrate-builder/download.zip" -o "substrate-builder.zip" && unzip -o "substrate-builder.zip" && find "substrate-builder" -name SKILL.md -print | while IFS= read -r file; do dir="$(dirname "$file")"; target="$HOME/.claude/skills/$(basename "$dir")"; mkdir -p "$target" && cp -R "$dir/." "$target/"; done

The skill

Define the decision, close ambiguous rules, and test consistent outputs across runs and models.

Substrate Builder
Converts a loose business prompt into a substrate, an explicit specification (definitions, thresholds, boundaries, precedence, closed labels, missing-data rules, canonical output) that makes different runs and different LLMs return the same answer, then proves it with an oracle coverage check and a cross-model drift test.

How to start

1. Download the full package and keep its folder structure intact.
2. Install the substrate-builder folder in your AI assistant's skills directory, or upload the full package where skills are supported.
3. Ask the assistant to build a substrate for a classification or scoring prompt. Approve the proposed decision rules before testing.
4. Use the included Python scripts and examples to check coverage and compare results across runs and models.

The individual SKILL.md describes the process; its tests and templates require the full package.
This package was supplied as a complete folder and is preserved in the download.

User-supplied Substrate Builder package. It reconstructs Elin Nguyen's interpretation-drift method and is not an official OmnisensAI tool.