Graph init
Initialize Graphify on an existing project. Builds the code knowledge graph, configures the Claude Code integration, installs git hooks for incremental rebuild, and adds graphify-out/ to .gitignore. Only useful for existing projects — greenfield projects have no ROI.From its SKILL.md
npx -y skills add RomainDECOSTER/ironforge --skill graph-initAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
4.0 KB, as published. Nobody here has run it
Ironforge Graph Init
Initialize Graphify on the current project. This creates a persistent code knowledge graph
that BMAD agents can query via the graph-explore subagent.
Step 1 — Check existing graph
ls graphify-out/ 2>/dev/null
If graphify-out/ exists, ask the user:
A Graphify graph already exists. Choose an option:
- Rebuild from scratch (
graphify update .— overwrites existing graph)- Cancel
Wait for the user's response. If cancel, stop.
Step 2 — Install Graphify
graphify --version 2>/dev/null || pip install graphifyy
If pip install fails, tell the user to run it manually and wait for confirmation before
continuing.
Step 3 — Platform setup (Claude Code integration)
graphify install
This configures Graphify for Claude Code:
- Writes the MCP server config to
.mcp.json - Installs a
PreToolUsehook insettings.json
Do not modify these files manually — Graphify owns them.
Step 4 — Build the graph
Ask the user:
How do you want to build the graph?
- Full pipeline (
/graphify .) — semantic extraction via LLM, community detection, HTML viz, Obsidian vault, GRAPH_REPORT.md. Costs tokens, takes a few minutes. Recommended for the first run.- Quick build (
graphify update .) — AST-only, no LLM, fast. No semantic edges, no community labels, no Obsidian vault.
Wait for the user's response.
If option 1: invoke the /graphify skill with the current directory as path (. or
the project root). The skill handles the full pipeline autonomously — do not duplicate its
steps here. Resume at Step 5 once it completes.
If option 2:
graphify update .
Output lands in graphify-out/:
graph.json— queryable graphgraph.html— visual explorer (open in browser)GRAPH_REPORT.md— top nodes and structural surprises
Note: Obsidian vault and semantic community labels are only generated by the full pipeline.
Run /graphify . --obsidian later to add them.
Step 5 — Always-on CLAUDE.md integration
graphify claude install
Adds a ## graphify section to the local CLAUDE.md that instructs Claude to check
the graph before answering codebase questions and rebuild it after code changes.
Step 6 — Install git hooks
graphify hook install
Installs post-commit and post-checkout hooks in .git/hooks/. These rebuild the graph
incrementally after every commit and branch switch, keeping the graph in sync without
manual intervention.
Step 7 — Update .gitignore
Check if graphify-out/ is already ignored:
grep -s "graphify-out" .gitignore
If not found, ask the user:
Add
graphify-out/to.gitignore? The graph is a local artefact — it should not be committed. (y/n)
If confirmed, append the line to .gitignore (create the file if it does not exist):
graphify-out/
Step 8 — Summary
Print the final status:
## Graphify — Initialized
✓ Graphify installed
✓ Claude Code integration configured (graphify install)
✓ Graph built (graphify-out/)
✓ Always-on CLAUDE.md integration (graphify claude install)
✓ Git hooks installed (post-commit, post-checkout)
✓ graphify-out/ ignored (.gitignore updated)
→ Open graphify-out/graph.html in a browser to explore the graph visually.
Open graphify-out/obsidian/ as a vault in Obsidian for graph view + canvas.
Run `/graphify . --obsidian` to regenerate with Obsidian vault if you used quick build.
BMAD agents will query the graph via the graph-explore subagent.