Sandbox
@backbay-labs/thrunt-god

Threat hunting workflow commands, agents, and hooks for agentic IDEs.

Thrunt God adds a structured hunt workflow to agentic IDEs through commands, agent definitions, and hooks. It gives you a repeatable path from signal to validated findings, with artifacts saved as files in a planning directory.

36 stars8 forksJavaScriptUpdated 2mo ago
Who it's for

Builders who run agentic IDEs for threat hunting, detection engineering, and incident analysis.

What it delivers

You can run a full hunt with shared commands, evidence, and validation instead of improvising each step.

What it does

Hunt commands

Provides commands like `/hunt:new-case`, `/hunt:map-environment`, `/hunt:plan`, `/hunt:run`, `/hunt:validate-findings`, and `/hunt:publish`.

Five-phase workflow

Structures each hunt into signal, hunt, swarm, receipt, and publish phases so work stays explicit.

Role-based agents

Includes many agent files such as `thrunt-hunt-planner`, `thrunt-query-writer`, `thrunt-evidence-correlator`, and `thrunt-findings-validator`.

Workflow hooks

Uses hooks like `thrunt-context-monitor`, `thrunt-prompt-guard`, and `thrunt-workflow-guard` to check behavior around agent actions.

Planning artifacts

Stores mission, hypotheses, queries, receipts, findings, and environment notes under `.planning/` or a custom planning directory.

Multi-IDE support

Works across Claude Code, Gemini, Codex, Copilot, Cursor, and Windsurf, with a VS Code extension alpha and MCP support.

How to get it

  1. 1Run
    npx thrunt-god@latest --claude --local
  2. 2From terminal
    code --install-extension thrunt-god-0.3.0.vsix

README

THRUNT GOD

npm CI Status Discord License: MIT

From signal, to swarm.
No gods. Only Thrunt.

Threat hunting command system for agentic IDEs.
Claude Code · OpenCode · Gemini · Codex · Copilot · Cursor · Windsurf

/thrunt:autonomous  |  one command, full hunt

Install  ·   Phases  ·   Commands  ·   Flows  ·   Artifacts


Installation

npx thrunt-god@latest --claude --local

thrunt-god install

Bootstrap the hunt command surface into your local IDE environment.

IDECommand
Claude Code / Gemini/hunt:help
OpenCode/hunt-help
Codex$hunt-help
Copilot/hunt-help
Cursor / Windsurfhunt-help

Install the VS Code extension alpha

Download thrunt-god-0.3.0.vsix from the v0.3.0 release.

Then install it one of two ways:

  1. In VS Code, open Extensions, click the ... menu, choose Install from VSIX..., and select the downloaded file.
  2. From terminal:
code --install-extension thrunt-god-0.3.0.vsix

After install, open a folder containing .planning/MISSION.md or .hunt/MISSION.md and the THRUNT sidebar will activate automatically.


The Five Phases

Every hunt resolves through five phases. Each step is explicit.

Signal → Hunt → Swarm → Receipt → Publish

Phase
SignalA detection, anomaly, lead, or intel input opens the case
HuntHypotheses are formed, scoped, and made testable
SwarmParallel agents execute structured investigations across available sources
ReceiptEvery claim is bound to exact queries, timestamps, and evidence lineage
PublishOnly validated findings are packaged for downstream consumers

Hunt Commands

CommandPurpose
/hunt:new-programStand up a long-lived hunt program
/hunt:new-caseOpen a case from a signal
/hunt:map-environmentInventory data sources, access, and topology
/hunt:shape-hypothesisDevelop and refine testable hypotheses
/hunt:plan <phase>Plan a hunt phase
/hunt:run <phase>Execute a hunt phase
/hunt:validate-findings [phase]Validate evidence chain for findings
/hunt:publish [target]Package and ship findings
/hunt:helpShow all commands and usage

Thrunt Commands

Utility and orchestration commands (/thrunt:*) for workspace management, diagnostics, settings, and agent control.


Common Flows

Single signal

/hunt:new-case
/hunt:shape-hypothesis
/hunt:plan 1
/hunt:run 1
/hunt:validate-findings 1
/hunt:publish

Long-lived program

/hunt:new-program
/hunt:map-environment
/hunt:new-case
  ... repeat per signal ...

Pack-seeded signal

/hunt:new-case --pack domain.identity-abuse
/hunt:run 1
/hunt:validate-findings 1

Autonomous

/thrunt:autonomous

Runs all remaining phases end-to-end: discuss, plan, execute. Pauses only for operator decisions.


Artifacts

All hunt state lives in a planning directory at the project root (.planning/ by default). Every query, receipt, and finding is a file, not a summary.

.planning/
├── config.json             # Project settings (mode, profile, connectors, workflow toggles)
├── MISSION.md              # Hunt program mission and scope
├── HYPOTHESES.md           # Testable hypotheses with status tracking
├── SUCCESS_CRITERIA.md     # Definition of done for the program
├── HUNTMAP.md              # Phase breakdown and execution roadmap
├── STATE.md                # Current phase, progress, blockers
├── FINDINGS.md             # Validated findings only
├── EVIDENCE_REVIEW.md      # Evidence chain audit
├── QUERIES/                # Exact queries run, with timestamps
├── RECEIPTS/               # Execution receipts per phase task
├── DETECTIONS/             # Detection rules promoted from findings
├── environment/
│   └── ENVIRONMENT.md      # Data source inventory and access map
├── phases/                 # Per-phase plans, research, and results
├── workstreams/            # Parallel hunt cases (optional)
├── milestones/             # Archived completed milestones
└── published/              # Final deliverables

Configuration

Settings live in .planning/config.json, created by /hunt:new-program and editable via /thrunt:settings. Global defaults in ~/.thrunt/defaults.json are merged into every new project config.

SettingDefaultWhat it controls
modeinteractiveinteractive confirms at each step, yolo auto-approves
granularitystandardPhase count: coarse (3-5), standard (5-8), fine (8-12)
model_profilebalancedModel tier per agent: quality, balanced, budget, inherit
planning.commit_docstrueWhether .planning/ is committed to git
git.branching_strategynonenone, phase (branch per phase), milestone (branch per version)

Full schema and connector profiles: docs/CONFIGURATION.md

Custom planning directory

Set THRUNT_PLANNING_DIR to change the directory name. This affects all path resolution, project root detection, and artifact storage.

export THRUNT_PLANNING_DIR=".hunt"

Storage

By default, .planning/ is committed to git so hunt artifacts travel with the repo. To keep artifacts local:

  1. Add .planning/ to .gitignore
  2. Set planning.commit_docs: false and planning.search_gitignored: true in config
  3. If previously tracked: git rm -r --cached .planning/

Workstreams (/thrunt:new-workspace) create isolated artifact trees under .planning/workstreams/{name}/ for parallel hunts in the same project.

Bootstrap fills confirmed fields immediately. TBD only marks live environment or operator-supplied facts that are still unknown.

Files in the repo

Repository payload27 top-level entries
  • .github
  • agents
  • apps
  • assets
  • bin
  • commands
  • docs
  • hooks
  • scripts
  • tests
  • thrunt-god
  • .base64scanignore
  • .gitignore
  • .release-monitor.sh
  • .secretscanignore
  • biome.json
  • CHANGELOG.md
  • CONTRIBUTING.md
  • LICENSE
  • package-lock.json
  • package.json
  • README.ja-JP.md
  • README.ko-KR.md
  • README.md
  • README.pt-BR.md
  • README.zh-CN.md
  • SECURITY.md

Discussion (0)

Ask about usage, or say what you built with it

Sign in to join the discussion.

No comments yet. Be the first to say what this is good for.

More harnesses

affaan-m/
ECC
affaan-m/ECCHarnesses

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

258k

The job search that runs on your machine. AI job application framework built on Claude Code: evaluate postings, tailor CVs, write cover letters, prep interviews. Fork it and own it.

42k
ruvnet/rufloHarnesses

🌊 The original agent meta-harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, RAG integration, and native Claude Code / Codex / Hermes and many more Integrated

72k

Practical patterns, starters & CLI tools for loop engineering with AI coding agents. Design systems that prompt and orchestrate agents (inspired by Addy Osmani and Boris Cherny). Includes loop-audit, loop-init, loop-cost.

11k