How do I install the Intake Normalizer skill?

Run npx skillmds@latest add richfrem/intake-normalizer in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Intake Normalizer safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Intake Normalizer?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Intake Normalizer free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Intake Normalizer?

richfrem (@richfrem) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Intake Normalizer by richfrem — Agent Skill

Name: Intake Normalizer
Author: richfrem

Intake Normalizer

Converts conversational input and images from the user into structured parameters. Handles missing or ambiguous data by defining defaults or returning flags for required clarification. Outputs validated design spec JSON to be ingested by the orchestrator.

richfrem 60bbe0d 28 files · 5.8 KB Updated Sep 17, 2026

File contents

Expected Inputs

Raw conversational transcript from the interactive-designer agent, and optionally paths to inspiration images.

Core Responsibilities

Parse the conversation transcript for exact structural intent (dimensions, structure type, site jurisdiction).
For missing parameters, inject safe structural defaults or mark them as NEEDS_CLARIFICATION.
Read the explicitly gathered building code jurisdiction (e.g., "British Columbia") and map it to a canonical location ID for the building-code-validator.
Output context/staging/structure.json with sections: meta, intent, structure, layout, presentation, cad. Set meta.lifecycle = "INTENT".

Gotchas

City ≠ jurisdiction. "Vancouver" could be Vancouver BC (Canadian codes) or Vancouver WA (US IBC). Always include province/state alongside city, or flag for clarification. Never resolve jurisdictions from city name alone.
Ambiguous footprints default to square. "About 12x12" with no shape specified implies a 4-post square structure, not hexagonal. Output the assumed shape explicitly in structure.json; do not silently assume.
Missing pitch defaults to 4:12 — log this default in the JSON as "pitch": "4:12", "pitchDefaulted": true so downstream agents can see the assumption.
NEEDS_CLARIFICATION blocks the pipeline. Any field marked NEEDS_CLARIFICATION will cause the structural engine to halt. Do not forward structure.json with unresolved clarifications unless the orchestrator has explicitly approved safe defaults.
Dimension vs. clear span. Users often give overall footprint (including post thickness). Structural span is the center-to-center post distance. Clarify which measurement the user intends when in doubt.

Smoke Test

Complete happy path: Given transcript "12-foot hexagonal gazebo in Kelowna, BC, 4:12 roof pitch": outputs structure.json with shape=hexagon, span_ft=12, jurisdiction=BC-Interior, pitch=4:12, zero NEEDS_CLARIFICATION fields. ✓
Missing data handling: Given transcript "I want a pergola, not sure about the size yet": outputs structure.json with NEEDS_CLARIFICATION on span_ft, pipeline does not advance. ✓
Cross-border jurisdiction: Given "Vancouver" with no province/state: outputs NEEDS_CLARIFICATION on jurisdiction rather than assuming BC. ✓

---name: intake-normalizerdescription: Converts conversational input and images from the user into structured parameters. Handles missing or ambiguous data by defining defaults or returning flags for required clarification. Outputs validated design spec JSON to be ingested by the orchestrator.---## Expected InputsRaw conversational transcript from the `interactive-designer` agent, and optionally paths to inspiration images.## Core Responsibilities1. Parse the conversation transcript for exact structural intent (dimensions, structure type, site jurisdiction).2. For missing parameters, inject safe structural defaults or mark them as `NEEDS_CLARIFICATION`.3. Read the explicitly gathered building code jurisdiction (e.g., "British Columbia") and map it to a canonical location ID for the `building-code-validator`.4. Output `context/staging/structure.json` with sections: meta, intent, structure, layout, presentation, cad. Set `meta.lifecycle = "INTENT"`.## Gotchas- **City ≠ jurisdiction.** "Vancouver" could be Vancouver BC (Canadian codes) or Vancouver WA (US IBC). Always include province/state alongside city, or flag for clarification. Never resolve jurisdictions from city name alone.- **Ambiguous footprints default to square.** "About 12x12" with no shape specified implies a 4-post square structure, not hexagonal. Output the assumed shape explicitly in structure.json; do not silently assume.- **Missing pitch defaults to 4:12** — log this default in the JSON as `"pitch": "4:12", "pitchDefaulted": true` so downstream agents can see the assumption.- **NEEDS_CLARIFICATION blocks the pipeline.** Any field marked `NEEDS_CLARIFICATION` will cause the structural engine to halt. Do not forward structure.json with unresolved clarifications unless the orchestrator has explicitly approved safe defaults.- **Dimension vs. clear span.** Users often give overall footprint (including post thickness). Structural span is the center-to-center post distance. Clarify which measurement the user intends when in doubt.## Smoke Test1. **Complete happy path:** Given transcript "12-foot hexagonal gazebo in Kelowna, BC, 4:12 roof pitch": outputs `structure.json` with `shape=hexagon`, `span_ft=12`, `jurisdiction=BC-Interior`, `pitch=4:12`, zero `NEEDS_CLARIFICATION` fields. ✓2. **Missing data handling:** Given transcript "I want a pergola, not sure about the size yet": outputs `structure.json` with `NEEDS_CLARIFICATION` on `span_ft`, pipeline does not advance. ✓3. **Cross-border jurisdiction:** Given "Vancouver" with no province/state: outputs `NEEDS_CLARIFICATION` on jurisdiction rather than assuming BC. ✓

Frequently asked questions