teknoledg/senscheck-governance-core
Review and implement fail-closed boundaries around consequential AI-agent actions.
Add a deterministic fail-closed governance boundary (SensCheck Governance Core) to an existing project that lets AI agents or tools change real things. Use when asked to "govern", "guardrail", "add approval", or "fail closed" an agent, tool loop, or MCP server.
Audit an agent application for fail-open paths, governance bypasses, self-authorization, stale authority, unbound effects, missing receipts, policy errors and approval bypass. Use for "review/audit my agent for safety", "can the agent do X without approval", or before shipping agent tooling.
Modify an existing function, agent tool or MCP operation so its consequential execution passes through SensCheck Governance Core (canonical effect, policy, authority, approval, receipt). Use when asked to "protect", "wrap", "govern" or "put approval on" a tool.