Add DomainSAFE

Claude Code plugin that generates individualized knowledge systems from conversation. You describe how you think and work, have a conversation and get a complete second brain as markdown files you own.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: add-domain
description: Add a new knowledge domain to your existing system. Derives domain-specific configuration through conversation, generates domain folders, templates, and vocabulary while preserving and connecting to your existing architecture.
context: fork
model: opus
allowed-tools: Read, Write, Edit, Grep, Glob, Bash, AskUserQuestion
argument-hint: "[domain name or description, e.g. 'therapy sessions' or 'creative writing']"
---

You are extending an existing knowledge system with a new domain. This is composition, not replacement. The new domain must coexist with existing domains while maintaining its own vocabulary, schema, and processing patterns. The shared graph (wiki links, hub MOC, description fields) connects everything.

## Your Task

Add a new knowledge domain: **$ARGUMENTS**

## Reference Files

Read these during derivation phases:

**Composition rules:**
- `${CLAUDE_PLUGIN_ROOT}/reference/derivation-validation.md` -- Test 4 (Multi-Domain Composition) validates the pattern
- `${CLAUDE_PLUGIN_ROOT}/reference/three-spaces.md` -- three-space architecture (shared across domains)
- `${CLAUDE_PLUGIN_ROOT}/reference/interaction-constraints.md` -- coherence validation for domain config

**Domain configuration:**
- `${CLAUDE_PLUGIN_ROOT}/reference/vocabulary-transforms.md` -- domain-native term mapping
- `${CLAUDE_PLUGIN_ROOT}/reference/tradition-presets.md` -- pre-validated domain configurations
- `${CLAUDE_PLUGIN_ROOT}/reference/use-case-presets.md` -- 3 presets with configurations
- `${CLAUDE_PLUGIN_ROOT}/reference/dimension-claim-map.md` -- research backing for dimension positions
- `${CLAUDE_PLUGIN_ROOT}/reference/failure-modes.md` -- domain vulnerability matrix

**Validation:**
- `${CLAUDE_PLUGIN_ROOT}/reference/kernel.yaml` -- the 12 non-negotiable primitives
- `${CLAUDE_PLUGIN_ROOT}/reference/validate-kernel.sh` -- kernel validation script

---

## PHASE 1: Scan Existing System

Automated. Understand what exists before adding to it.

### 1a. Read system configuration

Read `ops/derivation.md` for:
- Current dimension positions
- Vocabulary mapping
- Platform and automation level
- Active feature blocks

Read `ops/config.yaml` for live configuration values.

### 1b. Inventory existing domains

Identify all current knowledge domains:
- Primary notes folder and its domain vocabulary
- Any existing secondary domains (folders, templates, MOCs)
- Schema fields in use per domain

### 1c. Identify dimension types

Classify each of the 8 dimensions as system-level or domain-adjustable:

| Dimension | Type | Rationale |
|-----------|------|-----------|
| Organization | system-level | Flat/hierarchical applies to the whole workspace |
| Automation | system-level | Hooks and skills are workspace-wide infrastructure |
| Navigation depth | system-level | Hub MOC structure is shared |
| Granularity | domain-adjustable | Different domains may need different granularity |
| Processing | domain-adjustable | Research needs heavy, relationships need light |
| Maintenance | domain-adjustable | Different condition thresholds per domain growth rate |
| Schema | domain-adjustable | Domain-specific fields vary |
| Linking | domain-adjustable | Some domains need semantic search, others don't |

System-level dimensions are fixed by the existing system. Domain-adjustable dimensions can vary for the new domain.

---

## PHASE 2: Conversation

1-3 conversation turns to understand the new domain. Use AskUserQuestion for each turn.

### Opening question

Ask ONE focused question about the new domain:

**"Tell me about [domain hint from $ARGUMENTS] -- what kinds of things will you track, and how does this relate to your existing [current domain vocabulary] work?"**

The second half is critical: understanding the relationship between domains drives composition decisions.

### Signal extraction

As the user responds, extract signals for domain-adjustable dimensions:

| Signal | Dimension | Position |
|--------|-----------|----------|
| "Quick notes about people" | Granularity | moderate |
| "Deep analysis of sessions" | Processing | heavy |
| "Just remember key moments" | Processing | light |
| "I revisit and update often" | Maintenance | tight thresholds |
| "Mostly static once captured" | Maintenance | lax thresholds |
| "Need to find patterns across entries" | Linking | explicit+implicit |

Also extract:
- **Volume estimate** -- how many notes per processing batch
- **Temporal dynamics** -- how fast does content change
- **Vocabulary** -- the user's own words for notes, processes, organization
- **Cross-domain relationship** -- how this connects to existing domain(s)

### Follow-up strategy

After the opening response, ask 1-2 follow-ups targeting:
1. Vocabulary confirmation -- "When you say [user's word], do you mean individual insights or longer entries?"
2. Cross-domain connection patterns -- "Will [new domain] notes connect to your [existing domain] notes? How?"

Do NOT ask about dimensions directly. Listen for them in natural conversation.

---

## PHASE 3: Derive Domain Configuration

### 3a. Map signals to domain-adjustable dimensions

For each adjustable dimension, determine the position for the new domain:
- User signals (highest priority)
- Closest use-case preset from `${CLAUDE_PLUGIN_ROOT}/reference/use-case-presets.md`
- Cascade from system-level dimensions

### 3b. Build vocabulary mapping

Read `${CLAUDE_PLUGIN_ROOT}/reference/vocabulary-transforms.md` for the transformation table.

Priority order:
1. User's own words from conversation
2. Use-case preset vocabulary
3. Closest reference domain blend

Build the complete mapping for the new domain:

| Universal Term | New Domain Term | Source |
|---------------|-----------------|--------|
| note | [term] | [user / preset / blend] |
| extract / reduce | [term] | [user / preset / blend] |
| connect / reflect | [term] | [user / preset / blend] |
| MOC | [term] | [user / preset / blend] |
| description | [term] | [user / preset / bl

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.