execution-control

Execution control layer for AI agents — prevents duplicate or incorrect real-world actions under retries, uncertainty, and stale context.

Local-first execution-control reference runtime for AI agent builders.

AI Decision Validator (AI Execution Firewall) Validates AI-generated decisions before execution using deterministic tests, fail-closed validation, and cryptographic audit.

Workflow manager for coffea HEP analyses - DAG-based execution, caching, chunk-level fault tolerance, and one-line switching between local, coffea-casa, and lxplus.

Deterministic Event-driven trading decision engine for shared Backtesting and Live semantics.

PFC separates the time authority from the signer so a compromised enforcement point can't silently backdate its own approvals — and makes replay and reordering detectable through monotonic, hash-chained receipts.

Python runtime of the Spine Lite kernel. Policy-gated execution + receipts for AI agents. MIT.

Execution control engine for critical infrastructure. Docker-deployable. Fail-closed by default.

Official Python SDK for MachineID - simple device registration and validation for AI agents.

Ensures execution only occurs from the current committed state.

Reproducible consequence-boundary benchmark for agentic AI execution — same model proposal, different authority, measured consequence.

Evidence-bound execution control for autonomous product-analytics agents.

External kill switch for autonomous runtimes. Validate at enforcement boundaries. Revoke to halt execution.

Deterministic execution-control protocol for AI-orchestrated systems. Prevents execution of computational operations unless cryptographically authorized via state-transition artifacts. Specification-only repository.

Controlled execution for AI-agent tool calls. A reference runtime that routes every proposed tool call to ACCEPT, VERIFY, ABSTAIN or ESCALATE through a hash-pinned REMORA core, binds approval to the exact payload, and verifies the effect against the system of record.