This is built on ISG Agent 1, a governance-focused AI agent platform with constitution enforcement, adversarial self-testing, and cryptographic audit trails. The legal agent variant wraps that framework for contract review, clause drafting, compliance checklists, and legal research tasks. The underlying platform includes time-locked actions for dangerous operations, a trust ledger for reputation tracking, and skill quarantine for third-party integrations. If you need an AI to handle legal document workflows with built-in governance gates and explainable decision-making, this combines LLM reasoning with safety checks that log every action in a SHA-256 hash chain. Runs locally on stdio transport, localhost by default.
Governance from Day 1. A security-hardened, governance-first autonomous AI agent platform.
ISG Agent 1 proves that autonomous AI agents can be both powerful and safe. While other agent platforms bolt on security as an afterthought, ISG Agent 1 bakes governance into every layer from the start.
| Innovation | What It Does |
|---|---|
| Agent Constitution | Machine-enforced behavioral contract -- not guidelines, a verified contract |
| Adversarial Self-Testing | The agent red-teams itself in production on a schedule |
| Time-Locked Actions | Mandatory cooling period before dangerous operations (30-60s) |
| Trust Ledger | Transparent, cryptographic reputation tracking for every action |
| Explain Mode | Cryptographic proof of why every decision was made |
| Skill Reputation | Community-verified trust scores for agent skills |
| Separation of Powers | Critical actions require approval from independent agent or human |
ISG Agent 1 was built as a direct response to the security failures in existing agent platforms. Where others have exposed instances, malicious skills, and no audit trails, ISG Agent 1 has localhost-only defaults, skill quarantine, and hash-chained audit logs.
git clone https://github.com/InnovativeSystemsGlobal/isg-agent-1.git
cd isg-agent-1
cd gateway
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp ../.env.example ../.env
cp ../config/agent.example.yaml ../config/agent.yaml
python -m isg_agent
The gateway starts on http://localhost:8900 by default (localhost-only).
User (Discord/Telegram/Web)
-> Bridge (TypeScript, normalizes messages)
-> Gateway (Python/FastAPI, governance engine)
-> Constitution check
-> Governance gate (PROCEED/REVIEW/HALT)
-> Audit trail (SHA-256 hash chain)
-> Brain (LLM + convergence guarantees)
-> Skills (sandboxed, quarantined, reputation-scored)
-> Response (governed, explained, audited)
See ARCHITECTURE.md for the full architecture overview.
See CONTRIBUTING.md for contribution guidelines.
See SECURITY.md for our responsible disclosure policy.
MIT License. See LICENSE for details.
Innovative Systems Global. The name is not aspirational. It is a statement of fact.
DINGDAWG_API_KEYsecretAPI key for paid tier access — get free at dingdawg.com
io.github.ericm1018/skillfm-llm-cost-optimizer-openai-anthropic-usage
io.github.mikerawsonnz/llm-orchestration-agent
io.github.mikerawsonnz/authenticated-llm-agent
labforgedev/copilot-memory-mcp
csoai-org/agent-prompt-injection-firewall-mcp
io.github.mikerawsonnz/authenticated-multi-llm-agent