By haingt-dev
Core hooks — brain-powered memory injection, Semantic Toolbox, safety, entity extraction. Includes staleness warnings, FTS5 pre-filter, circuit breaker, session-count gate, and 9-section PreCompact snapshot.
A personal infrastructure layer for Claude Code — centralizing plugins, skills, hooks, and project bootstrapping across a multi-project workspace.
Claude Code is powerful but each project is an island. Configuration, skills, and memory are project-local by default. This hub creates a shared layer: global skills available everywhere, plugins that inject context at session start, a registry that tracks all projects, and templates for bootstrapping new ones.
/alfred (life scheduler), /mentor, /gen-image, /story, /token-optimizehaint-core (session hooks, context injection), godot-dev (Godot workflows)Note: All agent rules and configs are per-project. This hub only contains shared scripts, templates, and plugins.
~/Projects/agent/
├── bootstrap-project.sh # Bootstrap new project + auto-register in hub
├── ag-sync-rules.sh # Sync shared rules to child projects + update registry
├── ag-registry-audit.sh # Full drift check: registry vs actual state
├── registry.json # Reverse index of all child projects
├── shell-aliases.sh # Shell shortcuts (source in ~/.zshrc)
├── .claude/scripts/
│ └── registry-check.sh # SessionStart hook: lightweight drift alert
├── plugins/
│ ├── haint-core/ # Core plugin: hooks (SessionStart, PreToolUse, Notification)
│ └── godot-dev/ # Godot plugin: gdformat, GDScript workflows
└── templates/
├── memory-bank/ # Templates for new project Memory Banks
│ └── stories/ # Story template + index
├── agents/ # Sub-agent templates (code-reviewer, security-reviewer)
├── .env.example
└── .gitignore-secrets
Every project has this structure (created by bootstrap):
project/
├── .memory-bank/ # Project knowledge (brief, product, context, task, arch, tech)
│ └── stories/ # Dev stories for devlogs (not auto-loaded)
├── .claude/
│ ├── CLAUDE.md # Project context, values, memory bank, security
│ ├── settings.json # Project-specific hooks (if any)
│ └── skills/<name>/ # Skills (SKILL.md + supporting files)
└── .mcp.json # Project-level MCP servers (where needed)
Rules are minimized to reduce per-turn token cost:
| What | Where | Token cost |
|---|---|---|
| Enforcement (security, dangerous commands) | settings.json hooks | 0 (runs as shell) |
| Core directives (no dirty state, reversibility) | .claude/CLAUDE.md Values | Once per session |
| Memory Bank context | SessionStart hook output | Once per session |
| Project-specific workflows | skills/<name>/SKILL.md | On invocation only |
.claude/skills/<name>/SKILL.md)Skills use YAML frontmatter for invocation control:
/namesettings.json)| Hook | Purpose |
|---|---|
SessionStart | Inject git status + Memory Bank context |
PreToolUse (Bash) | Block dangerous commands, scan for secrets in staged files |
Stop (project-specific) | Auto-format on save (e.g., gdformat for Godot) |
The hub maintains a reverse index (registry.json) of all child projects — which plugins, rules, skills, and MCPs each project uses.
| Component | Role |
|---|---|
registry.json | Source of truth: project metadata, resources used |
ag-registry-audit.sh | Full audit: compares registry against filesystem + installed_plugins.json |
.claude/scripts/registry-check.sh | SessionStart hook: silent when clean, alerts on drift |
bootstrap-project.sh | Auto-registers new projects on bootstrap |
ag-sync-rules.sh | Updates registry rules after syncing |
Drift detection: hybrid pull model — hub SessionStart detects drift automatically, scripts that modify children update the registry as a side effect.
ag-help # Show all commands
bootstrap <dir> # Setup new project
ag-status # Check agent setup across all projects
mbk / mbc / mbt # Edit Memory Bank / context.md / task.md
cdc <project> # Switch to project directory
MIT License
Executes bash commands
Hook triggers when Bash tool is used
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
npx claudepluginhub haingt-dev/agent --plugin haint-coreGodot 4.x debugging and GDScript patterns — corrective lens, project conventions, and on-demand references for Wildtide and Chimera Protocol
Harness for Claude Code — skills, /harness:* slash commands, persona subagents, lifecycle hooks, and MCP tools without per-repo `harness setup`. Sibling plugins exist for Cursor, Gemini CLI, and Codex.
Unified capability management center for Skills, Agents, and Commands.
Persistent memory for Claude Code sessions using Honcho
HelloAGENTS — The orchestration kernel that makes any AI CLI smarter. Adds intelligent routing, unified QA gates, safety guards, and notifications.
Persistent memory, shared standards, and structured workflows for AI coding agents. Detects project setup and injects agent context automatically.
Complete collection of battle-tested Claude Code configs from an Anthropic hackathon winner - agents, skills, hooks, and rules evolved over 10+ months of intensive daily use