Asi magic-todo

Add TODOs to org files and generate ADHD-friendly task breakdowns with local MLX models on Apple Silicon. Load when creating actionable checklists from vague tasks or populating org-mode files with structured TODO entries.

install

source · Clone the upstream repo

git clone https://github.com/plurigrid/asi

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/plurigrid/asi "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/magic-todo" ~/.claude/skills/plurigrid-asi-magic-todo && rm -rf "$T"

manifest: skills/magic-todo/SKILL.md

source content

magic-todo

Setup

Python:
```
/Users/alice/v/.venv-mlx-lm/bin/python
```

Script:

/Users/alice/v/scripts/magic_todo_mlx.py

Default model:
```
mlx-community/Qwen3-8B-4bit
```

Elisp:

/Users/alice/v/magic-todo-org/emacs/magic-todo-org.el

CLI Usage

unset PYTHONPATH

# Generate JSON breakdown
echo "your task here" | /Users/alice/v/.venv-mlx-lm/bin/python \
  /Users/alice/v/scripts/magic_todo_mlx.py \
  --format json --spice 2 \
  --model mlx-community/Qwen3-8B-4bit \
  --max-tokens 800

# Generate org-mode checklist
echo "your task here" | /Users/alice/v/.venv-mlx-lm/bin/python \
  /Users/alice/v/scripts/magic_todo_mlx.py \
  --format md --spice 2 \
  --model mlx-community/Qwen3-8B-4bit \
  --max-tokens 800

Org File Entry Format

* TODO <task description>
:PROPERTIES:
:MAGIC_TODO_TASK: <task description>
:MAGIC_TODO_SPICE: <1-5>
:MAGIC_TODO_MODEL: mlx-community/Qwen3-8B-4bit
:END:
- [ ] step one
- [ ] step two

Spice Levels

Level	Meaning	Guidance
1	Trivial	2-3 quick steps
2	Easy	4-6 clear steps
3	Medium	6-10 steps, may need substeps
4	Hard	8-15 steps with substeps
5	Overwhelming	10-20 steps, deeply broken down

Parameters

Flag	Default	Description
`--format`	`text`	Output format: `text` , `md` , `json`
`--spice`	`2`	Difficulty 1-5
`--model`	`Qwen3-8B-4bit`	MLX model repo id
`--max-tokens`	`2048`	Max generation tokens
`--temp`	`0.7`	Sampling temperature
`--top-p`	`0.9`	Nucleus sampling
`--context`	-	Existing org content for style learning
`--list-models`	-	List cached MLX models

Workflow

Run the script with
```
--format json
```
to get structured output

Parse the JSON:

{"title": "...", "steps": [{"text": "...", "substeps": null}]}

Format as org checklist (
```
- [ ] step text
```
)
Append to the target
```
.org
```
file with PROPERTIES block

Output Rules

Steps are concise (~10 words max), verb-first
Tools are always suggested with alternatives ("using X or Y")
No filler: no "review", "test", "document", "audit", "ensure", "plan"
ADHD-friendly: many small steps, no walls of text

Known Org Files

/Users/alice/vi/plurigrid.org/plurigrid-basinator.org