AGENT SKILL

graphify-code-graph

Deterministic, queryable knowledge graph over a codebase with graphify (tree-sitter, 25+ languages, no LLM cost). Use it for code-level impact analysis inside a client repo, where grep answers poorly.

License 许可
MIT
最近更新
2026-11-14
Stars
★ 18
Forks
6

01 前言 Front Matter

name
graphify-code-graph
metadata
{ "type": "tool-skill", "source": "https://github.com/Graphify-Labs/graphify. This skill DECLARES no license and its manifest carries the repo default over ORIGINAL text: measured at the entry ref b1936079 against all 361 markdown files upstream, tree not truncated: max similarity 0.0519 (0.2091 with difflib autojunk off) and longest shared run 43 characters. An earlier note here said 40 files, 0.0502 and 21 characters; that was a truncated 11% sample published as a census, and the conclusion survived the correction while the receipt did not, so nothing here is copied and this repo's own terms are the right ones. The TOOL was MIT when this was written and upstream relicensed to Apache-2.0 on 2026-07-22 (ba7f9ea9), six days after this entered on 2026-07-16 (a14b67c). That is recorded so nobody 'corrects' the manifest against the current listing: the tool's license was never what this field carried.", "deep-learn": "knowledge/repo-deep-learn/graphify/2026-07-16.md" }
description
Deterministic, queryable knowledge graph over a codebase with graphify (tree-sitter, 25+ languages, no LLM cost). Use it for code-level impact analysis inside a client repo, where grep answers poorly.

02 资源文件 1

03 SKILL.md 正文

Graphify: deterministic code knowledge graph

When to use

  • A code question spans many files ("what breaks if I change X?", "how does A reach B?") and grep gives partial, noisy answers.
  • Onboarding into a large or unfamiliar codebase (client repo, OSS dependency).
  • Repeated architecture questions in one repo: build the graph once, query it every session at near-zero token cost.

Not for concept/governance recall. Which skill knows this, where does a rule live, what derives from a doc: that is the brain's connectome (query_connectome.py) and lineage (impact-radius.py). Graphify maps code symbols; those map concepts. Two layers, keep both.

Install (one time per machine)

uv tool install graphifyy      # or: pipx install graphifyy
graphify install               # registers the /graphify skill with the AI assistant

The PyPI package is graphifyy (double y); the CLI is graphify.

Build the graph (inside the target repo)

graphify .          # writes graphify-out/graph.json + GRAPH_REPORT.md + graph.html
graphify update .   # incremental re-extract of changed paths only, AST-only, no API cost
graphify watch      # keep it fresh on file changes

The code pass is fully deterministic: tree-sitter AST, no LLM, no API key, $0. Only the optional semantic pass (docs, PDFs, images, video, community labels) calls an LLM; without a key it is skipped entirely. Add graphify-out/ to the repo's .gitignore unless the team wants the artifact versioned.

Query it

graphify query "how does auth reach the billing module"
graphify affected src/auth/session.py    # code-level impact radius
graphify path NodeA NodeB                # shortest path between symbols
graphify explain "PaymentProcessor"      # natural-language node explanation
graphify diagnose                        # graph health

Matching is case-folded substring + IDF over the flat graph.json (NetworkX node-link format). No server, no embeddings. For assistant integration without shell calls there is an MCP server: graphify-mcp (stdio/HTTP).

Arm usage pattern

Run it inside ONE arm's repo; the graph lives in that repo's graphify-out/ and never leaves the arm (arm isolation applies to the artifact like to any other client data). Start every code-impact question with graphify affected <file> and fall back to grep only for symbols the graph does not know, same seek-before-scan discipline as the brain's graphs.

Gotchas

  • Edges carry confidence (EXTRACTED / INFERRED / AMBIGUOUS, 0.55-0.95 rubric): trust EXTRACTED, verify INFERRED before acting on it.
  • The exporter refuses to silently shrink an existing graph (their issue #479). The brain adopted the same invariant in scripts/generate_neural_map.py (SHRINK-GUARD).
  • Exports available when a human wants to browse: Obsidian vault, GraphML (Gephi/yEd), SVG, Cypher (Neo4j/FalkorDB).

See also

  • [[repomix]]: flat single-file codebase packing, the non-graph alternative for one-shot context stuffing.
  • [[4d-paradigm-protocol]]: graphify affected is the code-level twin of the concept-level Impact Radius scan.

04 同仓库更多 Skills

CarlosCaPe
CarlosCaPe/octorato

yeet

Use only when the user explicitly asks to stage, commit, push, and open a GitHub pull request in one flow using the GitHub CLI (`gh`).

★ 18
CarlosCaPe
CarlosCaPe/octorato

wrangler

Cloudflare Workers CLI for deploying, developing, and managing Workers, KV, R2, D1, Vectorize, Hyperdrive, Workers AI, Containers, Queues, Workflows, Pipelines, and Secrets Store. Load before running wrangler commands to ensure correct syntax and best practices. Biases towards retrieval from Cloudflare docs over pre-trained knowledge.

★ 18
CarlosCaPe
CarlosCaPe/octorato

workspace-skill-discovery

Discover and include ALL skills across the workspace — not just global ones. Use at the start of every session or when the user asks about available skills. Ensures project-level skills (under .claude/skills/ or skills/ in any repo) are never missed.

★ 18
CarlosCaPe
CarlosCaPe/octorato

workers-best-practices

Reviews and authors Cloudflare Workers code against production best practices. Load when writing new Workers, reviewing Worker code, configuring wrangler.jsonc, or checking for common Workers anti-patterns (streaming, floating promises, global state, secrets, bindings, observability). Biases towards retrieval from Cloudflare docs over pre-trained knowledge.

★ 18
CarlosCaPe
CarlosCaPe/octorato

winui-app

Create, build and design WinUI 3 desktop apps with C# and the Windows App SDK following Microsoft's official guidance and CommunityToolkit. Covers XAML, navigation, theming, accessibility and deployment.

★ 18
CarlosCaPe
CarlosCaPe/octorato

windows-brain-script-portability

Scripts in ~/.claude/scripts/ that run on a Windows brain: use the python3.cmd shim and not bare python3, and reconfigure stdout to UTF-8 if you print symbols. Use it when writing a brain script or triaging a Windows-only crash.

★ 18