ΣΛ (Sigma–Lambda)

🌐 sigmalambda.dev

ΣΛ is a constraint protocol for safe, auditable AI systems. It preserves human intent under automation by separating:

• Policy (what must be true / what must never happen / when to stop)
• Procedure (how work is executed: agents, scripts, CI/CD)
• Evidence (what actually happened: verifiable traces)

ΣΛ is not a programming language and does not execute workflows. It is the governance layer that makes systems stoppable, auditable, and defensible.

Why ΣΛ

As AI agents gain autonomy and persistence, failures increasingly come from:

• inferred intent
• silent scope expansion
• “helpful” optimization past safety boundaries
• lack of explicit halting semantics

ΣΛ makes limits explicit and enforceable.

Repository structure

• spec/ — Canonical specification and governance
• proposals/ — ΣΛ Change Proposals (SΛCP)
• examples/ — Minimal examples (policy / ops / trace)
• tools/ — Reference validator (sl_validate.py)
• templates/ — System prompts for AI-assisted authoring
• guides/ — Authoring and validation workflows
• docs/ — Public-facing materials (whitepaper, one-pager)

Start here

1. Read the Operating Manual: spec/SL_Operating_Manual_v1.1.md
2. Write your first policy: guides/authoring_workflow.md
3. Use the system prompt: templates/system_prompt_v1.md
4. See a worked example: examples/agent/

License

Apache-2.0 (see LICENSE)