Hivemind Graph
AST code graph you query as files — callers, imports, impact, no build
Test report
- Verdict
- Works with setup
- Score
- Tested
- Jul 20, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Aug 10, 2026 · d17651a
Fetched the repo tree and the raw Claude Code SKILL.md plus plugin.json/marketplace.json. The skill is clean prose (allowed-tools: Read Bash) that `cat`s paths like ~/.deeplake/memory/graph/query/<pattern> — no scripts, curl|sh, base64, or exfiltration. OUTPUT is NOT measured (not a measured loss): the graph mount only exists after `npm i -g @deeplake/hivemind && hivemind install` wires hooks + a virtual filesystem + Deeplake sign-in, an external CLI I don't install; copying the SKILL.md alone leaves it inert (docked 1 install point). Docs are unusually honest — explicit "When NOT to use", staleness caveats, and an anti-pattern warning that "Incoming (0)" isn't dead code. TRIGGER phrasings — SHOULD load: (1) "What calls pushSnapshot in this repo?" (2) "If I change this function's signature, what's the blast radius?" (3) "What are the main subsystems / the architecture of this codebase?"; should NOT: (4) "Fix the failing test in auth.test.ts" (5) "Write a new function to parse CSV" — all 5 judged correctly.
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 4/5
- Triggers reliably 5/5
- Output vs. baseline 5/10
- Docs & honesty 5/5
What Hivemind Graph does
A pure-instruction skill that answers structural codebase questions (what calls X, what imports Y, where is Z defined, change impact) by reading an AST-derived graph exposed as synthesized files under the Deeplake mount. It triggers on relational/architecture questions and points the agent to source_file:line to read the real code. It is inert until the external @deeplake/hivemind CLI is installed and the mount is populated by its background hooks.
How to install Hivemind Graph
npm i -g @deeplake/hivemind && hivemind install (then the Claude Code plugin skill loads; graph mount auto-builds via hooks)
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Hivemind Graph
-
/hivemind-graphAST code graph you query as files — callers, imports, impact, no build
It also activates on plain-language prompts like these:
-
What calls this function across the codebase -
What would break if I change this module -
Show me the architecture of this subsystem
Frequently asked questions
- Is the Hivemind Graph skill free?
- Yes. The skill itself is free from activeloopai/hivemind. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Hivemind Graph work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 20, 2026. Verdict: Works with setup. Fetched the repo tree and the raw Claude Code SKILL.md plus plugin.json/marketplace.json. The skill is clean prose (allowed-tools: Read Bash) that `cat`s paths like ~/.deeplake/memory/graph/query/<pattern> — no scripts, curl|sh, base64, or exfiltration. OUTPUT is NOT measured (not a measured loss): the graph mount only exists after `npm i -g @deeplake/hivemind && hivemind install` wires hooks + a virtual filesystem + Deeplake sign-in, an external CLI I don't install; copying the SKILL.md alone leaves it inert (docked 1 install point). Docs are unusually honest — explicit "When NOT to use", staleness caveats, and an anti-pattern warning that "Incoming (0)" isn't dead code. TRIGGER phrasings — SHOULD load: (1) "What calls pushSnapshot in this repo?" (2) "If I change this function's signature, what's the blast radius?" (3) "What are the main subsystems / the architecture of this codebase?"; should NOT: (4) "Fix the failing test in auth.test.ts" (5) "Write a new function to parse CSV" — all 5 judged correctly.
- What is the Hivemind Graph SkillProof Score?
- 7.6/10 — installs cleanly 4/5, triggers reliably 5/5, output vs. baseline 5/10, docs & honesty 5/5.
- How do I install Hivemind Graph?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Hivemind Graph with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.