Persistent context for Claude Code.

Skills that establish and maintain long-term knowldege without filling your context window.

GitHub automation that understands your org.

Projects v2, milestones, branch protection — all require hunting for opaque GraphQL node IDs. hiivmind-pulse-gh discovers your org structure once, caches it, and works in plain English from then on.

/hiivmind-pulse-gh create issue for login timeout bug
/hiivmind-pulse-gh set milestone v2.0 on issue #42
/hiivmind-pulse-gh add PR to project
/hiivmind-pulse-gh what operations do I have permission to perform on this repo?

No more PVT_kwDOBx... or PVTSSF_... hunting.

What it provides:

• 6 skills — init, refresh, operations, discover, awareness, plus GitHub docs corpus
• 25 domain libraries — Issues, PRs, Projects v2, Milestones, Labels, Branch Protection, Rulesets, Actions, Secrets, Variables, Releases, and more
• Discovery mode — explore all available operations across domains with interactive drill-down
• Corpus-backed syntax — automatic GraphQL/REST lookup when you need exact API syntax

Persistent documentation indexes for any project.

Think of it like Claude Projects for Claude Code, but better. With Projects, you dump docs, code, and PDFs into a collection that persists across chats. But everything consumes context, there's no way to prioritize what matters, no freshness tracking, no namespacing, no search.

A corpus solves all of this. You build a curated index once, Claude searches it directly without filling your context window, and the index tracks exactly where everything came from and how fresh it is. Install as many as you need. They're lightweight and independent.

This implements the "just in time" context pattern recommended by Anthropic: lightweight identifiers that load content on demand, like how humans use indexes rather than memorizing corpuses.

/hiivmind-corpus

One command, natural language:

"Create a corpus for Polars"
"How do lazy frames work?"
"Add the TanStack Query docs to my fullstack corpus"
"Is my React corpus up to date?"

Combine git repos, local documents, and web articles into unified corpora. Eight skills handle the full lifecycle—discover, navigate, init, add-source, build, enhance, refresh, upgrade.

Pre-built documentation indexes, ready to install.

Data engineering libraries:

Claude ecosystem:

GitHub platform:

Complete GitHub documentation — Actions, Copilot, REST API, GraphQL, code security, repositories, and more. Used by hiivmind-pulse-gh for API syntax lookup.

Discover once, use forever. Structure is stable; data changes constantly. Cache the IDs, fields, and mappings — not the content.

Skills over MCP. Load on-demand, not all upfront. Better context efficiency for complex workflows.

Human-readable everything. YAML configs, markdown indexes. If a human can't read it, a human can't fix it.

Install via the Claude Code marketplace:

# GitHub automation
/plugin marketplace add hiivmind/hiivmind-pulse-gh
/plugin install hiivmind-pulse-gh@hiivmind-pulse-gh

# Documentation corpus tooling
/plugin marketplace add hiivmind/hiivmind-corpus
/plugin install hiivmind-corpus@hiivmind-corpus

# Pre-built corpora (install what you need)
/plugin marketplace add hiivmind/hiivmind-corpus-data
/plugin install hiivmind-corpus-polars

/plugin marketplace add hiivmind/hiivmind-corpus-claude
/plugin install hiivmind-corpus-claude-agent-sdk

/plugin marketplace add hiivmind/hiivmind-corpus-github
/plugin install hiivmind-corpus-github-docs

_{Melbourne, AU}