I created a SKILL.md file for IWE

[kiil]

UPDATE - I attached the skill for ease.

SKILL.md

Hi - thanks again for IWE and also the recent additions making it even stronger when also working with LLM based workflows and tools.

I made a SKILL.md very much based upon the excellent documentation at the IWE site for it that seems to work well, at least for me. I tested it with claude code and Gemini CLI:

---

---

name: iwe
description: Traverse, search, retrieve, and manage documents within the IWE note collection and knowledge graph. Use when you need to find notes, extract sections, inline references, analyze document hierarchy, or modify the knowledge system.

IWE CLI Commands Reference

Use the iwe CLI to interact with the local note collection and knowledge graph.

iwe find

Discover documents via fuzzy search on keys and titles.

iwe find [QUERY] [OPTIONS]

Flag	Purpose	Default
--roots	Only show root documents (no parents)	false
--refs-to <KEY>	Documents that reference KEY	—
--refs-from <KEY>	Documents that KEY references	—
-f, --format	markdown, keys, json	markdown

• No query: lists all documents ranked by incoming references (popularity)
• Results capped at 50

iwe find                          # All docs by popularity
iwe find "AI strategy"            # Fuzzy search
iwe find --roots -f keys          # Entry points, pipeable
iwe find --refs-to my-doc         # What links to my-doc

iwe retrieve

Fetch document content with graph expansion (children, parents, links). Documents are deduplicated.

iwe retrieve [OPTIONS]

Flag	Purpose	Default
-k, --key <KEY>	Document key(s) to retrieve	stdin
-d, --depth <N>	Follow children N levels deep	1
-c, --context <N>	Include N parent levels	1
-l, --links	Include inline references	false
-b, --backlinks	Include incoming references	true
-f, --format	markdown, keys, json	markdown
-e, --exclude <KEY>	Exclude specific keys	—
--no-content	Metadata only	false
--dry-run	Show stats without content	false

iwe retrieve -k my-doc                   # Basic retrieval
iwe retrieve -k my-doc -d 3 -c 2         # Deep expansion
iwe retrieve -k my-doc --no-content      # Metadata only
iwe retrieve -k my-doc --dry-run         # Check size first
# Exclude already-loaded docs to avoid duplication
iwe retrieve -k topic-b $(iwe retrieve -k topic-a -f keys | xargs -I {} echo "-e {}")

iwe tree

Show the navigable document hierarchy.

iwe tree [OPTIONS]

Flag	Purpose	Default
-k, --key <KEY>	Starting document(s); repeatable	—
-d, --depth <N>	Traversal depth limit	4
-f, --format	markdown, keys, json	markdown
-v, --verbose <N>	Detail level (1=info, 2=debug)	0

iwe tree                          # Full hierarchy
iwe tree -k my-doc --depth 2      # Limited from a root
iwe tree -f json                  # Structured output
iwe tree | grep -i api            # Filter results

iwe stats

Generate knowledge base statistics and analytics.

iwe stats [OPTIONS]

Flag	Purpose	Default
-f, --format	markdown, csv	markdown

iwe stats                         # Full report
iwe stats -f csv > stats.csv      # Export for analysis
iwe stats -f csv | cut -d, -f1,2,3 | column -t -s,

CSV columns: key, title, sections, paragraphs, lines, words, parents, incoming_inline_refs, total_incoming_refs, children, outgoing_inline_refs, total_connections, bullet_lists, ordered_lists, code_blocks, tables, quotes

iwe extract

Extract a section from a document into a new standalone document. Replaces original section with an inclusion link. Resolves key collisions with numeric suffix.

iwe extract <KEY> [OPTIONS]

Flag	Purpose
--section <TITLE>	Section to extract (case-insensitive)
--block <N>	Section by block number (1-indexed)
--list	List all sections with numbers
--action <NAME>	Use named action from config
--dry-run	Preview without writing
--quiet	Suppress progress output
--keys	Output affected keys

iwe extract my-doc --list                      # See all sections
iwe extract my-doc --section "Configuration"   # Extract by title
iwe extract my-doc --block 3                   # Extract by position
iwe extract my-doc --section "Design" --dry-run

iwe inline

Embed a referenced document's content directly into the parent document. Replaces inclusion link with actual content. Source document deleted by default unless --keep-target.

iwe inline <KEY> [OPTIONS]

Flag	Purpose
--reference <KEY>	Reference key/title to inline
--block <N>	Select by block number (1-indexed)
--list	Show all inclusion links with numbers
--as-quote	Wrap content in blockquote
--keep-target	Preserve source document after inlining
--dry-run	Preview without writing
--keys	Output affected keys
--action <NAME>	Use named action from config
--quiet	Suppress progress output

iwe inline notes/index --list
iwe inline notes/index --reference "architecture"
iwe inline notes/index --block 1
iwe inline notes/index --reference "design" --dry-run
iwe inline notes/index --block 2 --keep-target
iwe inline notes/index --reference "notes" --as-quote

iwe delete

Remove a document and clean up all references to it across the knowledge base. Removes inclusion links pointing to the deleted doc and converts inline links to plain text (preserving link text).

iwe delete <KEY> [OPTIONS]

Flag	Purpose
--dry-run	Preview changes without writing
--force	Skip confirmation prompt
--quiet	Suppress progress output
--keys	Output affected keys

iwe delete old-notes                      # With confirmation
iwe delete obsolete-doc --dry-run         # Preview first
iwe delete temp-note --force              # No prompt
iwe delete my-doc --keys --dry-run        # Scripting preview
iwe delete doc --quiet --force            # Silent automation

iwe rename

Rename a document and update all references to it. Updates both inclusion links and inline references. Preserves directory structure for path-based keys.

iwe rename <OLD_KEY> <NEW_KEY> [OPTIONS]

Flag	Purpose
--dry-run	Preview without modifying files
--quiet	Suppress progress output
--keys	Output affected keys line-by-line

iwe rename my-note renamed-note
iwe rename old-key new-key --dry-run
iwe rename document new-document --keys

Typical Agentic Workflow

# 1. Discover entry points
iwe find --roots -f keys

# 2. Check scope before loading
iwe retrieve -k topic --dry-run

# 3. Load content with context
iwe retrieve -k topic -d 2 -c 1

# 4. Avoid re-loading already seen docs
iwe retrieve -k related-topic -e already-seen-key

# 5. Understand full structure
iwe tree -k topic -f json

If you want to use it in the project or base something else on it, please do.

[gimalay]

Thanks for sharing @kiil ! How did you create the SKILL, what was the prompt?

[kiil]

I can't remember the exact prompt. But I pointed claude code at the relevant documentation page and it seems to have found the correct subpages by itself.

The first SKILL made by claude did not work in Gemini, though. But Gemini managed to make some changes to make it work and now it works in both.

[gimalay]

iwe command has build in detailed documentation for every sub-command. You can ask Claude/Codex to learn the command and it can do it by using iwe --help. This is how I create SKILL's for myself.

[kiil]

yes - I noticed that the command documentation is very good, even with examples included :)

[gimalay]

Closing in favor of #244

Please feel free to contribute in the skills repo.