Subagent Driven Writing logo

Subagent Driven Writing

Community
binbinao
subagent-driven-writing

Fast iterative writing using fresh subagent per section with two-stage review. Alternative to batch execution.

Overview

Publisherbinbinao
Repositorydocument-superpowers
Skill namesubagent-driven-writing
Stars
49
Forks
14
Bundled files
Instructions only
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by binbinao on GitHub. Read the source before you install it.

Installation

Install the Subagent Driven Writing AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/binbinao/document-superpowers/tree/main/skills/subagent-driven-writing
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/binbinao/document-superpowers.git /tmp/document-superpowers
mkdir -p .claude/skills
cp -r /tmp/document-superpowers/skills/subagent-driven-writing .claude/skills/subagent-driven-writing
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Subagent Driven Writing in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Subagent Driven Writing on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Subagent Driven Writing is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Subagent-Driven Writing

Overview

Dispatch a fresh subagent for each article section, review their output, iterate if needed, then continue. Faster than batch execution when you want tight feedback loops.

Announce at start: "I'm using subagent-driven-writing for fast iterative writing."

Context: Should follow an approved plan from writing-article-plan.

Two-stage review per section:

  1. Spec compliance - Does it follow the plan?
  2. Content quality - Is the writing good?

When to Use This

Use subagent-driven-writing when:

  • You want tight control over each section
  • The article is complex or high-stakes
  • You want to iterate quickly on feedback
  • Sections are interdependent (later sections depend on earlier ones)

Use executing-article-plan instead when:

  • You want to write in focused batches
  • The article is straightforward
  • Sections are largely independent
  • You prefer fewer but longer review sessions

The Process

dot
digraph subagent {
    "Read plan" [shape=box];
    "Spawn subagent for section N" [shape=box];
    "Subagent writes section" [shape=box];
    "Review 1: Spec compliance" [shape=diamond];
    "Review 2: Content quality" [shape=diamond];
    "Save section" [shape=box];
    "More sections?" [shape=diamond];
    "Merge into full draft" [shape=box];
    "Invoke reviewing-article" [shape=doublecircle];

    "Read plan" -> "Spawn subagent for section N";
    "Spawn subagent for section N" -> "Subagent writes section";
    "Subagent writes section" -> "Review 1: Spec compliance";
    "Review 1: Spec compliance" -> "Spawn subagent for section N" [label="fail, retry"];
    "Review 1: Spec compliance" -> "Review 2: Content quality" [label="pass"];
    "Review 2: Content quality" -> "Spawn subagent for section N" [label="needs revision"];
    "Review 2: Content quality" -> "Save section" [label="pass"];
    "Save section" -> "More sections?";
    "More sections?" -> "Spawn subagent for section N" [label="yes"];
    "More sections?" -> "Merge into full draft" [label="no"];
    "Merge into full draft" -> "Invoke reviewing-article";
}

Spawning Subagents

For each section in the plan:

markdown
**Spawning subagent for Section N: [Title]**

Task brief:
- Read plan from: docs/writing/YYYY-MM-DD-<topic>-plan.md
- Focus on: Section N ([Title])
- Write section following plan specifications
- Target word count: XXX
- Maintain tone: [from plan]
- Include all key points
- Save to: docs/writing/drafts/YYYY-MM-DD-<topic>/section-N-<title>.md

Use sessions_spawn:

Write Section N ([Title]) for article [Topic].
Read the plan from docs/writing/YYYY-MM-DD-<topic>-plan.md.
Follow the specifications for Section N exactly:
- Purpose: [from plan]
- Key points: [list]
- Target word count: XXX
- Tone: [style]
Include smooth transition from Section N-1.
Save output to docs/writing/drafts/YYYY-MM-DD-<topic>/section-N-<title>.md with metadata header.

Review Stage 1: Spec Compliance

Check against plan:

  • All key points covered?
  • Word count within range (±10%)?
  • Supporting evidence included?
  • Tone matches guidelines?
  • Transition from previous section present?
  • Transition to next section present?

If fails:

  • Identify specific gaps
  • Spawn new subagent with clarified requirements
  • Include what was missing in new task brief

Example failure:

markdown
**Review failed: Spec compliance**

**Missing:**
- Key point 2 (example of enterprise use case)
- Transition to next section

**Retrying with updated brief...**

Review Stage 2: Content Quality

Check writing quality:

  • Logic clear and sound?
  • Evidence supports claims?
  • Paragraphs well-structured (2-4 sentences)?
  • Sentence variety (not all same length/structure)?
  • Active voice used where appropriate?
  • No jargon without explanation?
  • Smooth reading experience?

If needs revision:

  • Identify specific issues
  • Spawn new subagent with feedback
  • Include what needs improvement

Example revision:

markdown
**Review: Content quality needs improvement**

**Issues:**
1. Paragraph 2 uses passive voice heavily
2. Example in paragraph 3 doesn't clearly support the point
3. Transition to next section is abrupt

**Revising...**

If passes:

  • Save section with approval note
  • Continue to next section

Section Approval Format

markdown
## Section N Review: ✅ APPROVED

**Spec Compliance:** PASS
- All key points covered ✓
- Word count: XXX (target: XXX, within range) ✓
- Evidence included ✓
- Tone matches ✓
- Transitions present ✓

**Content Quality:** PASS
- Logic clear ✓
- Evidence strong ✓
- Well-structured ✓
- Smooth reading ✓

**Saved to:** docs/writing/drafts/YYYY-MM-DD-<topic>/section-N-<title>.md

**Continuing to Section N+1...**

Handling Interdependencies

When sections depend on each other:

markdown
**Note:** Section 4 references the example from Section 2.

**Subagent task includes:**
- Read Section 2 draft before writing
- Reference the [specific example] when discussing [topic]
- Maintain consistency with terminology used in Section 2

Context passing:

  • Each subagent gets the plan + all completed sections
  • Explicitly note dependencies in task brief
  • Include specific cross-references needed

Progress Tracking

After each section:

markdown
## Progress: Section N/M Complete

**Completed:**
- ✅ Section 1: Introduction (152 words)
- ✅ Section 2: First Argument (298 words)
- ✅ Section 3: Second Argument (305 words)

**Current:**
- 🔄 Section 4: Third Argument (in progress)

**Remaining:**
- ⏳ Section 5: Conclusion

**Stats:**
- Current word count: 755
- Target word count: 1200
- Progress: 63%

After All Sections

Merge and transition:

markdown
**All sections complete! 🎉**

**Section Summary:**
- Section 1: ✅ (152 words)
- Section 2: ✅ (298 words)
- Section 3: ✅ (305 words)
- Section 4: ✅ (289 words)
- Section 5: ✅ (163 words)

**Total:** 1207 words (target: 1200)

**Merging into full draft...**

Saved to: docs/writing/YYYY-MM-DD-<topic>-draft.md

**Next:** Invoke reviewing-article for systematic review

Then:

  • REQUIRED SUB-SKILL: Use document-superpowers:reviewing-article

Key Differences from Batch Execution

Subagent-DrivenBatch Execution
Fresh subagent per sectionOne agent writes all sections
Review after each sectionReview every 2-3 sections
Tight feedback loopFewer interruptions
Better for complex articlesBetter for simple articles
More overhead (spawning)Less overhead
Easier to course-correctMore focused flow state

Troubleshooting

If subagent consistently fails spec compliance:

  • Plan may be unclear - revise plan
  • Requirements may be contradictory - clarify plan
  • Word count may be unrealistic - adjust target

If subagent produces low-quality content:

  • Add more specific tone/style examples
  • Include writing samples in task brief
  • Specify common pitfalls to avoid

If reviews take too long:

  • Create review checklist template
  • Focus on critical issues only
  • Consider switching to batch execution

Key Principles

  • Fresh subagent = fresh perspective - Each section gets unbiased writing
  • Two-stage review = quality control - Spec first, then quality
  • Pass context explicitly - Don't assume subagent knows prior sections
  • Iterate quickly - Fast feedback beats batch-and-hope
  • Track progress visibly - Know where you are in the process
  • Don't over-review - Good enough to move forward is good enough

Frequently asked questions

What does the Subagent Driven Writing AI skill do?

Fast iterative writing using fresh subagent per section with two-stage review. Alternative to batch execution.

Why use Subagent Driven Writing on TypingMind?

Because you install it once and use it with any model. Subagent Driven Writing is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Subagent Driven Writing in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/binbinao/document-superpowers/tree/main/skills/subagent-driven-writing. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Subagent Driven Writing?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Subagent Driven Writing?

As many as you like. As long as a model supports skills, you can use Subagent Driven Writing with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Subagent Driven Writing AI skill free?

It is published on GitHub by binbinao. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇