Skip to main content

Agent Knowledge Contract

What must an external AI agent know before it can trust, call, or improve Dreamineering?

The Spine​

  • Open Knowledge Format Mapping — tells agents why markdown, frontmatter, links, and logs make knowledge portable.
  • Delegation — defines the decision-rights layer for trusted autonomy.
  • Agent Operating Model — names identity, capability contracts, machine-readable surfaces, and trust signals.
  • DDL Nomenclature — defines the human-readable canonical language for agents, instruments, receipts, loops, and decisions.
  • Dreamineering Symbols — defines the compressed DML and A&ID notation agents use as wire-format shorthand.
  • Language Crosswalk — binds DDL terms to DML codes, A&ID symbols, color roles, and semantic layers.
  • Actions And Consequences — keeps agent-facing artifacts tied to one reader, one tension, one dream, and one first action.
  • MCP Toolkit — separates MCP reality, desired tool posture, classification, and human-owned remediation.
  • Agent Funnel — preserves the agent discovery path from manifest to call, payment, and receipt.
  • Wiki Schema — defines the read order, page jobs, and source/projection rule for agent-facing knowledge.
  • Naming Standards — keeps routes, terms, skills, and generated projections unambiguous.
  • Agent Systems — compares AI-agent frameworks and ecosystems.
  • Stackmates — proves which platform services and outcomes are real.

Zoom Out​

This folder is the canonical source for agent-facing knowledge contracts. The old public /agents pages were retired so source truth cannot drift across two surfaces. Durable agent knowledge belongs in OKF-readable playbook pages that other AI agents can quote, link, export, and improve.

The contract has six jobs. It must tell an agent what language to use, what capability it can call, what decision rights it has, what is proven, what should be built next, and how the loop improves. Capability claims should name REALITY, DREAM, or CONSUMED status so a reader can separate built proof from desired demand.

The migration order recovered the old /agents route into OKF-readable source pages: DDL nomenclature, Dreamineering symbols, language crosswalk, actions and consequences, MCP toolkit, agent operating model, and agent funnel. Each move ends with one source-of-truth home, not a rendered duplicate.

The lock-in layer is usage. Knowledge is durable only when skills, hooks, loops, Stackmates proof, and receipts make the right action easier next time.

Delegation is the bridge from readable knowledge to trusted action. The contract should make act-alone, ask-first, escalation, and refusal boundaries obvious before an agent touches tools.

Questions​

Which source contract would let the next agent act with less translation?

  • Which old agent page still has source truth trapped in code data?
  • Which capability lacks REALITY, DREAM, or CONSUMED status?
  • Which service outcome can Stackmates prove today?
  • Which hook should prevent the next route or naming drift?

Context​

  • Purpose — why the agent contract exists before any capability claim.
  • Principles — the truths that decide whether an agent action should run.
  • Platform — where reusable capability becomes operational substrate.
  • Performance — the proof layer for REALITY, DREAM, and CONSUMED status.