
Helps agents and developers implement Agent Control Contracts (ACC) well. Provides a comprehensive template, where to place ACC in the architecture, and how ACC differs from AGENTS.md. Use when defining workflow governance, policy envelopes, or linking AGENTS.md to governance.
--- name: aaf-security description: Applies the AAF Security pillar: boundaries, tool actuation, epistemic gatekeeping, supply chain. Use when designing security for agentic systems, tool gateways, prompt-injection mitigations, privilege separation, or supply-chain risk for skills/tools. --- # AAF Security Security in agentic systems is impact-reduction: assume the reasoning layer can be influenced, and architect so that influence cannot easily become harmful actions. This skill distills the A
--- name: aaf-cost-context description: Applies AAF Cost pillar and context optimization: budgets, model routing, token economics, context discipline. Use when controlling cost in agentic systems, budgeting context, choosing model routing, or designing for token economics. --- # AAF Cost & Context Cost optimization in agentic systems is an architectural requirement: autonomy without budgets is cost volatility by design. Context optimization cross-cuts cost, performance, reliability, and securi
Helps build agent orchestration properly using the Orchestrator Capability Contract (OCC) and governance-above-orchestration pattern. Use when choosing or implementing orchestration (graphs, multi-agent, workflows), ensuring tool gateway is non-bypassable, or satisfying OCC requirements for audit and safety.
Applies the Agentic Architecture Framework when designing or reviewing agentic systems. Covers pillar checklist, trade-offs, design-time spec and pre-production review. Use when designing agentic systems, reviewing architectures, pre-production readiness, or discussing autonomy and governance trade-offs.
Applies the two AAF cross-cutting foundations—Context Optimization and Autonomy & Outcome Governance. Use when designing context vs memory, context budgeting, autonomy levels, Definition of Done, budgets, or escalation.
Guides when and where to place epistemic gates; candidate → validated → authority; gates scale with risk. Use when defining validation vs authority, reducing "AI said so" authority, or designing high-stakes decision flows.