Content
**im-not-ai v2.0 + Process Engineer (PE) Style Extension**
[](LICENSE)
> Original: [epoko77-ai/im-not-ai](https://github.com/epoko77-ai-ai) (MIT)
> Extension: [dreamworker0/ur-socialworker](https://github.com/dreamworker0/urimal-for-socialworker) (MIT)
---
is this?
This is a tool for refining technical reports and analysis documents **without AI jargon, in a Process Engineer style**.
| Before | After |
|---------|---------|
| "⭐ In conclusion, this section confirms." | "[Key] This section confirms." |
| H1~H5 section descriptions (6, 1200 characters) | 1 hypothesis comparison table + 2 diagnosis principles (2 |
| "~ occurs", "~ is essential" | "~ occurs", "~ is essential" |
| Descriptive prose | Imperative bullet points (`▶`)-centric |
---
## What's new compared to the original im-not-ai.0
### 5 PE (Process Engineer) rules
| Description | Effect |
|------|------|------|
| PE-1 compression | Delete or summarize explanations engineers already know → | Reduce content by 20~30% |
| PE-2 Explanatory text → Imperative | "~ does this because" → `▶` bullet points | Improve readability |
| PE-3 Hypothesis section → Summary table | H1~H5 individual descriptions → 1 hypothesis comparison table | Reduce content by 40% |
| PE-4 Verb conjugation | "confirms" → "confirms it" | Engineer writing style |
| PE-5 Content goal | Original content compressed by 40~50% | Focus on key points |
### Items to preserve (absolutely not deleted even with PE rules)
- Numerical values, specifications, and criteria (e.g., PCE 18.2%, n=12)
- Entire tables
- Decision Gate rules
- Risk response plans
- Measurement item lists
---
## MCPConnect to Claude Code and call `mcp__cng-ui__doc_humanize` directly Installation (based on cng-ui MCP)
```bash
# 1. Copy humanize.py to the cng-ui MCP directory
cp/humanize.py //cng-ui-mcp/
# 2. Import and register the tool in server.py (refer to mcp/server_patch.py)
# or refer to the "server.py patch" section in README
```
### server.py Patch
```python
# Add to the import section at the top of server.py
import humanize as HZ
# Add the following @mcp.tool() block
@mcp.tool()
def doc_humanize(content: str, str = "v20pe_path: str = "") -> str:
""" documents in im-not2.0 + Process Engineer (PE) style.
file_path: If a file provided, read the file instead of content.
mode: 'v20pe' (default) | 'v20' (refining only) | 'pe' (PE compression only)
Returns (JSON):
applied — List of applied machine rules
content — applying machine rules
guidance — Additional judgment rules for Claude to apply
"""
if:
from pathlib import Path
p = Path(file_path if not p.exists():
import json
return jsonerror": f"File not found: {file_path}"},=False)
content =_text(encoding="utf- if not content.strip import json
return json.dumps({"error": "content or required"}, ensure_ascii=False import json
result = HZ.humanize(content, mode=mode)
result["modes"] = HZ.MODES
json.dumps(result, ensure_ascii=False, indent=2)
Usage
```python
# Code
mcp__cng-ui__doc_human file_path="/var/www_report.html",
mode20pe" # | pe | v20pe
)
```
The `content` return value contains the text rules applied, and `guidance` contains additional rules for Claude.
---
## Claude Code
```bash
# the Claude Code skill directory
cp -r .claude/skills/humanize-pe ~/.claude/skills/
```
It will be activated in subsequent conversations with `/human` or related prompts.
---
Virtual scenario: Per thin film spin coating uniformity improvement (PSC-A small module)
- draft**: Emoji, pivot words, translation tone, noun H1~H3 detailed descriptions
- **v2.0 S1·S2 pattern removal, honorific conversion, pronoun correction
- **PE modified**: Content compressed to ~55%, hypothesis comparison table integrated, imperative bullet points converted
| Example | Link------|------|
| Before (AI draft) | [beforehttps://yazzangab.github.io/im-not-ai-pe/examples/before.html) |
| After (v2.0 + PE) | [after\_pe.html](https://yazzang-homelab.github.io/im-not-ai-pe/examples/after_pe.html) |
| Original·v2.0·v2.0+PE 3-stage comparison | [diff\_v20\_vs\_original.html](https://yazzang-homelab.github.io/im-pe/examples/diff_v_original.html) |
---
Summary (v2. PE integrated)
See [`-rules.md`](rules/pe-rules.md) for the complete rules Machine-automatically applied (humanize.py)
| Rule | Description | Severity |
|------|------|--------|
| C-5 |ojis and convert text | S1 |
| D-1 | Delete conclusion pivots (e.g., "in conclusion", "therefore") | S1 |
| C-11 | Remove commas after conjunctions | S1 |
| Translation tone | "this section" → "this section", "this case" → "this example", etc. | S1 |
| Emphasis symbols | ⭐→[Key], ★→[Priority 1] | S1 |
### Claude judgment application (delivered as guidance)
| Rule | Description | Severity |
|------|------|--------|
| Verb completion | Noun endings → honorific endings (~20 cases) | S1 |
| A-15 | Estimation → assertion ("~ will be" → "~ is") | S2 |
| A-16 | Pronoun direct translation correction (e.g., "it" → specific noun) | S1 PE-1~5 | Explanation compression, imperative, hypothesis | PE |
---
## Grade Criteria (v2.0)
| Grade | Conditions |
|------|------|
| **A** | S1=0, S2≤2, change rate 10~25% |
| B1≤2, S2≤5, change rate 10~30% |
| C1≤5 or change rate 30~50% |
| D | S1 6+, change rate 50%+ |
---
## License
MIT License — original author attribution required:
-77-ai (im-not-ai)
- dreamworker0 (urimal-for-socialworker)
See [LICENSE](LICENSE) for details.
