jmagly/aiwg/doc-splitter

Claude Code and agent skill by jmagly/aiwg

Trust tier 2
  • community
LLM review pending
Source check pending

Checking that the SKILL.md file is still in the repository.

Split large documentation (10K+ pages) into focused sub-skills with intelligent routing. Use for massive doc sites like Godot, AWS, or MSDN.

Install

npx vskill@latest install jmagly/aiwg/doc-splitter

SKILL.md

Documentation Splitter Skill

Purpose

Single responsibility: Split large documentation sites into multiple focused sub-skills with an optional router skill for intelligent navigation. (BP-4)

Grounding Checkpoint (Archetype 1 Mitigation)

Before executing, VERIFY:

  • [ ] Total page count is known (run estimation first)
  • [ ] Documentation categories are identifiable
  • [ ] Target skill size determined (default: 5,000 pages per skill)
  • [ ] Router strategy selected (category, size, or hybrid)

DO NOT split without understanding documentation structure.

Uncertainty Escalation (Archetype 2 Mitigation)

ASK USER instead of guessing when:

  • Category boundaries unclear
  • Optimal skill size uncertain for target use case
  • Cross-references between sections complicate splitting
  • Router vs flat structure decision needed

NEVER arbitrarily split - seek user guidance on boundaries.

Context Scope (Archetype 3 Mitigation)

Context TypeIncludedExcluded
RELEVANTDoc structure, categories, page countsActual page content
PERIPHERALSimilar large doc examplesOther documentation
DISTRACTORContent quality concernsIndividual page issues

Size Guidelines

Documentation SizeRecommendationStrategy
< 5,000 pagesOne skillNo splitting
5,000 - 10,000 pagesConsider splittingCategory-based
10,000 - 30,000 pagesRecommendedRouter + Categories
30,000+ pagesStrongly recommendedRouter + Categories

Workflow Steps

Step 1: Estimate Documentation Size (Grounding)
bash
# Quick estimation with skill-seekers
skill-seekers estimate configs/large-docs.json
# Output:
# πŸ“Š ESTIMATION RESULTS
# βœ… Pages Discovered: 28,450
# πŸ“ˆ Estimated Total: 32,000
# ⏱️ Time Elapsed: 2.1 minutes
# πŸ’‘ Recommended: Split into 6-7 sub-skills
Step 2: Analyze Category Structure
bash
# Identify natural category boundaries
skill-seekers analyze --config configs/large-docs.json --categories
# Output:
# Categories detected:
# - scripting: 8,200 pages
# - 2d: 5,400 pages
# - 3d: 9,100 pages
# - physics: 4,300 pages
# - networking: 2,800 pages
# - editor: 2,200 pages
Step 3: Choose Split Strategy
StrategyBest ForDescription
categoryClear topic divisionsSplit by documentation sections
sizeUniform distributionSplit every N pages
routerUser navigationHub skill + specialized sub-skills
hybridComplex docsCategories + size limits per category
Step 4: Execute Split

Option A: With skill-seekers

bash
# Category-based split
skill-seekers split --config configs/godot.json --strategy category
# Router-based split (recommended for large docs)
skill-seekers split --config configs/godot.json --strategy router
# Size-based split
skill-seekers split --config configs/godot.json --strategy size --pages-per-skill 5000

Option B: Manual split configuration

json
{
"name": "godot",
"max_pages": 40000,
"split_strategy": "router",
"split_config": {
"target_pages_per_skill": 5000,
"create_router": true,
"categories": {
"scripting": {
"patterns": ["/scripting/", "/gdscript/", "/c_sharp/"],
"max_pages": 8000
},
"2d": {
"patterns": ["/2d/", "/sprite/", "/tilemap/"],
"max_pages": 6000
},
"3d": {
"patterns": ["/3d/", "/mesh/", "/spatial/"],
"max_pages": 10000
},
"physics": {
"patterns": ["/physics/", "/collision/", "/rigidbody/"],
"max_pages": 5000
}
}
}
}
Step 5: Scrape Sub-Skills
bash
# Scrape all sub-skills in parallel
for config in configs/godot-*.json; do
skill-seekers scrape --config $config &
done
wait
# Or sequentially with progress
for config in configs/godot-*.json; do
echo "Processing: $config"
skill-seekers scrape --config $config
done
Step 6: Generate Router Skill
bash
# Auto-generate router from sub-skills
skill-seekers generate-router configs/godot-*.json
# Creates godot-router skill that intelligently routes queries
Step 7: Validate Split Results
bash
# Check sub-skill sizes
for dir in output/godot-*/; do
echo "$dir: $(find $dir -name "*.md" | wc -l) files"
done
# Verify router coverage
cat output/godot-router/SKILL.md | grep -A 50 "## Sub-Skills"

Recovery Protocol (Archetype 4 Mitigation)

On error:

  1. PAUSE - Note which sub-skill failed
  2. DIAGNOSE - Check error type:
  • Category overlap β†’ Refine URL patterns
  • Uneven split β†’ Adjust page limits
  • Orphan pages β†’ Add catch-all category
  • Router incomplete β†’ Regenerate after all sub-skills done
  1. ADAPT - Modify split configuration
  2. RETRY - Re-split affected category (max 3 attempts)
  3. ESCALATE - Present split preview, ask user for boundary adjustments

Checkpoint Support

State saved to: .aiwg/working/checkpoints/doc-splitter/

checkpoints/doc-splitter/
β”œβ”€β”€ estimation.json # Page count results
β”œβ”€β”€ category_analysis.json # Category breakdown
β”œβ”€β”€ split_plan.json # Planned split configuration
β”œβ”€β”€ progress/
β”‚ β”œβ”€β”€ godot-scripting.json
β”‚ β”œβ”€β”€ godot-2d.json
β”‚ └── ...
└── router_draft.md # Router skill draft

Output Structure

After splitting large documentation:

configs/
β”œβ”€β”€ godot.json # Original config
β”œβ”€β”€ godot-scripting.json # Generated sub-config
β”œβ”€β”€ godot-2d.json
β”œβ”€β”€ godot-3d.json
β”œβ”€β”€ godot-physics.json
└── godot-router.json # Router config
output/
β”œβ”€β”€ godot-scripting/ # Sub-skill
β”‚ β”œβ”€β”€ SKILL.md
β”‚ └── references/
β”œβ”€β”€ godot-2d/ # Sub-skill
β”œβ”€β”€ godot-3d/ # Sub-skill
β”œβ”€β”€ godot-physics/ # Sub-skill
└── godot-router/ # Router skill
β”œβ”€β”€ SKILL.md # Routing logic
└── references/
└── routing-table.md

Router Skill Structure

The generated router skill:

markdown
# Godot Documentation Router
## Purpose
Route queries to the appropriate specialized Godot sub-skill.
## Sub-Skills
| Topic | Skill | Coverage |
|-------|-------|----------|
| GDScript, C#, scripting patterns | godot-scripting | 8,200 pages |
| 2D graphics, sprites, tilemaps | godot-2d | 5,400 pages |
| 3D graphics, meshes, materials | godot-3d | 9,100 pages |
| Physics, collisions, rigid bodies | godot-physics | 4,300 pages |
## Routing Rules
1.**Scripting questions** β†’ godot-scripting
-Keywords: script, gdscript, c#, function, variable, class
2.**2D graphics questions** β†’ godot-2d
-Keywords: sprite, 2d, tilemap, animation2d, canvas
3.**3D graphics questions** β†’ godot-3d
-Keywords: mesh, 3d, spatial, material, shader, camera3d
4.**Physics questions** β†’ godot-physics
-Keywords: physics, collision, rigidbody, area, raycast
## Usage
Ask your question naturally. This router will direct you to the appropriate specialized skill.
Example:
-"How do I create a player movement script?" β†’ godot-scripting
-"How do I set up tilemap collisions?" β†’ godot-2d
-"How do I apply materials to a mesh?" β†’ godot-3d

Troubleshooting

IssueDiagnosisSolution
Uneven splitsCategory size variesUse hybrid strategy with max_pages
Orphan pagesURL patterns incompleteAdd catch-all or refine patterns
Router confusionOverlapping keywordsMake routing rules more specific
Too many skillsOver-segmentedMerge related categories

References

  • Skill Seekers Large Documentation: https://github.com/jmagly/Skill_Seekers/blob/main/docs/LARGE_DOCUMENTATION.md
  • REF-001: Production-Grade Agentic Workflows (BP-4, BP-9 KISS)
  • REF-002: LLM Failure Modes (Archetype 3 context filtering, Archetype 4 recovery)

Security

  • Security scanPassed, 96 of 100Pattern checks on the skill's files.
  • Model reviewNot runA model review is scheduled for the next rescan.

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
96
Repository forks
12
Repository npm downloads, 7 days
2,424
7-day trend score
41

Stars and forks are for the whole GitHub repository, not this skill alone. Downloads are for the repository's npm package. Refreshed 19 Mar 2026, 7 months ago.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/jmagly/aiwg/doc-splitter/badge)](https://verified-skill.com/skills/jmagly/aiwg/doc-splitter)

Something wrong with this skill or its page? Tell us.