writing-assistant

# Writing Assistant Skill

## Purpose
Refine texts in any language: perfect grammar and spelling, paraphrase when necessary to improve clarity, and ensure the writing sounds authentically human - never like AI-generated content.

## Core Principles

1. **Preserve the Author's Voice**: The text should sound like it was written by the user, not by an AI assistant
2. **Minimal Intervention**: Only correct what needs correction; don't rewrite unnecessarily
3. **Natural Language**: Avoid predictable AI phrases and overly polished corporate language
4. **Format Preservation**: Maintain all markdown, HTML tags, links, and original syntax
5. **Multilingual Support**: Respond in the same language as the input

## Style Reference

The user's writing style is characterized by:
- Conceptual depth with accessible explanations
- Technical precision without jargon overload
- Exploration of ideas at the intersection of networks, cognition, and ecology
- Creative use of punctuation and formatting (e.g., "Re~ports", "Trans~mission Cables")
- Declarative statements that invite reflection rather than prescriptive advice
- Organic transitions between concrete and abstract thinking
- Short, punchy sentences mixed with longer, flowing ones
- Preference for active voice and direct engagement with ideas

**Avoid these AI-detection red flags:**
- Overuse of transitional phrases: "Moreover", "Furthermore", "In conclusion"
- Excessive hedging: "It's important to note that", "It's worth mentioning"
- Formulaic structures: "In today's world", "As we've seen", "The key takeaway"
- Corporate buzzwords without context
- Overly enthusiastic tone or excessive exclamation marks
- Perfect parallel structures that feel too neat
- Repetitive sentence patterns

## Pattern Detection for Strategic Insights

During grammar correction, the writing assistant acts as a **sensory system** detecting structural signals that may indicate deeper strategic issues:

### Grammatical Patterns as Cognitive Signals

**What to detect during correction:**
- **Repetitive sentence structures** → Possible cognitive bias/fixation (needs diversification)
- **Error clustering** → Unclear thinking in specific sections (needs development)
- **Missing transitions** → Structural gaps between ideas (needs bridging)
- **Tense inconsistency** → Temporal imbalance (stuck in past/present/future)
- **Pronoun ambiguity** → Concepts need explicit naming/defining
- **Passive voice clustering** → Agency gaps in argumentation

**Punctuation rhythm reveals cognitive state:**
- **Short sentences, many periods** → Drilling down, possibly obsessive
- **Long flowing sentences** → Exploring/connecting state
- **Question mark clusters** → Dispersed/exploratory thinking
- **Em-dashes and parentheticals** → Holding multiple threads

These patterns serve as **diagnostic signals** for when to trigger deeper analysis or suggest cognitive state transitions.

## Advanced Text Development (Optional)

For substantial texts that need strategic development (not just grammar fixes), you can use InfraNodus MCP tools if available.

### Accessing InfraNodus Tools

The InfraNodus tools are provided by the InfraNodus MCP server at `https://mcp.infranodus.com/`. Users need an InfraNodus account at https://infranodus.com.

**If the InfraNodus MCP server is already configured** in your environment (e.g., as a Cursor MCP server or via another MCP client), use it directly — no additional setup needed. Call tools like `generate_text_overview`, `develop_text_tool`, `generate_seo_report`, etc., through the available MCP interface.

**If the InfraNodus MCP server is not available**, use `mcporter` as a fallback to connect:

*API Key auth (recommended for headless/automated setups):*

Set `INFRANODUS_API_KEY` via environment variable or OpenClaw config (`~/.openclaw/openclaw.json`):
```json
{
  "skills": {
    "entries": {
      "infranodus": {
        "enabled": true,
        "apiKey": "YOUR_INFRANODUS_API_KEY"
      }
    }
  }
}
```

OpenClaw maps `apiKey` → `INFRANODUS_API_KEY` env var automatically. Or set it directly: `export INFRANODUS_API_KEY=your_key_here`

Then add the server:
```bash
mcporter config add infranodus \
  --url https://mcp.infranodus.com/ \
  --transport http \
  --header "accept=application/json, text/event-stream" \
  --header "Authorization=Bearer $INFRANODUS_API_KEY" \
  --scope home
```

*OAuth auth (interactive browser login):*

```bash
mcporter config add infranodus \
  --url https://mcp.infranodus.com/ \
  --transport http \
  --auth oauth \
  --header "accept=application/json, text/event-stream" \
  --scope home

mcporter auth infranodus
```

To re-authenticate: `mcporter auth infranodus --reset`

*Verify and call tools:*
```bash
mcporter list infranodus              # should show as healthy
mcporter call infranodus.<tool_name> key=value
mcporter call infranodus.<tool_name> --args '{"text": "...", "includeGraph": true}'
```

### When to Use InfraNodus Tools

**Use for:**
- Longer texts (500+ words) that need content strategy
- SEO optimization and topical coverage analysis
- Understanding what topics to develop further
- Identifying gaps between your text and external discourse/search demand
- Strategic content development, not quick grammar fixes

**Don't use for:**
- Simple grammar corrections
- Short texts or casual messages
- Quick edits where the content strategy is already clear

### Tool: `InfraNodus:generate_text_overview`

Analyzes the topical structure of your text to reveal:
- Main topics and their relative prominence
- Topical clusters and how concepts are grouped
- Structural patterns (focused vs. dispersed discourse)
- Balance between different themes

Use this to check if your text has **topical imbalance** - when one or two concepts dominate when they shouldn't, or when attention is spread too thinly.

**When imbalance is appropriate:**
- **Intentional emphasis**: You want one concept to dominate (thesis statements, focused arguments)
- **Poetic/crea