Intelligence

L5. The agent is the primitive. A skill is how it works, a tool is how it reaches the world, and the two are never the same thing.

1 min read Updated Aug 12, 2026

Intelligence

L5. The layer that decides.

An agent decides what to do. It does not execute, and it does not persist. The layer below it does both.

L5 INTELLIGENCE — the agent spec, the skill-to-tool relation, and the one-hop rule
L5 INTELLIGENCE — the agent spec, the skill-to-tool relation, and the one-hop rule
AGENT — one versioned row, not a class
AGENT — one versioned row, not a class
identity · instructions · model
identity · instructions · model
skills it may load
skills it may load
tools it may call
tools it may call
context policy · memory policy · budget
context policy · memory policy · budget
SKILL IS NOT TOOL — a skill uses tools, a tool never uses a skill
SKILL IS NOT TOOL — a skill uses tools, a tool never uses a skill
Research skill
how the agent performs a capability
Research skillhow the agent performs a capability
web search tool
web search tool
browser tool
browser tool
company database tool
company database tool
ONE HOP — an agent calls a workflow, and a workflow calls steps
ONE HOP — an agent calls a workflow, and a workflow calls steps
Agent A
reads the intent, writes a plan
Agent Areads the intent, writes a plan
Workflow
deterministic coordination
Workflowdeterministic coordination
Run · agent B
Run · agent B
Run · agent C
Run · agent C
Agent A reads both results
Agent A reads both results
An agent never calls another agent.
It fans out to runs.
An agent never calls another agent.It fans out to runs.
LEGEND
LEGEND
entry — the agent that starts
entry — the agent that starts
costs an LLM call
costs an LLM call
what the design forbids
what the design forbids
ordinary component
ordinary component
a grouping, not a component
a grouping, not a component
Text is not SVG - cannot display
L5 INTELLIGENCE — the agent spec, the skill-to-tool relation, and the one-hop rule
L5 INTELLIGENCE — the agent spec, the skill-to-tool relation, and the one-hop rule
AGENT — one versioned row, not a class
AGENT — one versioned row, not a class
identity · instructions · model
identity · instructions · model
skills it may load
skills it may load
tools it may call
tools it may call
context policy · memory policy · budget
context policy · memory policy · budget
SKILL IS NOT TOOL — a skill uses tools, a tool never uses a skill
SKILL IS NOT TOOL — a skill uses tools, a tool never uses a skill
Research skill
how the agent performs a capability
Research skillhow the agent performs a capability
web search tool
web search tool
browser tool
browser tool
company database tool
company database tool
ONE HOP — an agent calls a workflow, and a workflow calls steps
ONE HOP — an agent calls a workflow, and a workflow calls steps
Agent A
reads the intent, writes a plan
Agent Areads the intent, writes a plan
Workflow
deterministic coordination
Workflowdeterministic coordination
Run · agent B
Run · agent B
Run · agent C
Run · agent C
Agent A reads both results
Agent A reads both results
An agent never calls another agent.
It fans out to runs.
An agent never calls another agent.It fans out to runs.
LEGEND
LEGEND
entry — the agent that starts
entry — the agent that starts
costs an LLM call
costs an LLM call
what the design forbids
what the design forbids
ordinary component
ordinary component
a grouping, not a component
a grouping, not a component
Text is not SVG - cannot display
What the agent row holds, the skill-to-tool relation, and the one-hop rule: an agent calls a workflow, and the workflow calls the runs.

Components

ComponentJob
AgentOne versioned specification of a decision-maker.
SkillOne reusable procedure an agent loads to perform a capability.
PlannerTurn an instruction into an ordered plan, or select an existing workflow.
ReasonerThe model loop. Read the brief, choose a tool, read the result, decide again.
DelegationHand a plan to the orchestration layer as one run.

The agent

An agent is a row, not a class. Eight fields define it.

TEXT
agent
  identity              name, role, owner
  instructions          the system prompt
  model                 provider and model id
  skills                the procedures it may load
  tools                 the capability list it may call
  context policy        what the context builder retrieves for it
  memory policy         what it may read and what it may propose
  budget                cost, tokens and tool calls per run

A change to any field is a new version. A run pins the version it started with, so a spec edit never changes a run in flight.

Skill is not tool

This is the distinction the layer exists to hold.

SkillTool
AnswersHow does the agent perform a capability?How does the agent reach the world?
Lives inL5, this layerL2
ShapeA procedure: instructions, examples, checksA contract: name, input schema, output schema
Side effectNone on its ownUsually yes

A skill uses tools. A tool never uses a skill.

TEXT
Research skill
  ├── web search tool
  ├── browser tool
  └── company database tool

Collapsing the two is the common failure. It produces a "research tool" that hides a prompt, cannot be governed by a policy on its action, and cannot be reused by a second skill.

Rules

  • One agentic hop. An agent calls a workflow. A workflow calls steps. An agent does not call another agent. Routing errors multiply through nested reasoning chains.
  • An agent holds no credential. It names a tool. The tool layer resolves the credential.
  • An agent writes no SQL, and calls no raw route. Both are tool calls.
  • An agent does not choose its own context. A context policy on the row drives a deterministic builder. A retrieval decision by the agent is a second hop.
  • An agent does not decide its own authorization. It asks for an action. The policy plane answers.
  • Fan out to runs, never to agents. An agent that must do ten things starts ten runs. It does not start ten agents.
  • The spec is provider-neutral. One adapter maps it to a model provider. Lock-in stays a one-file problem.

Agent collaboration

Two agents cooperate through the run tree, and never through a direct call.

TEXT
Agent A → run (workflow)
             ├── step 1 → run (agent B)
             ├── step 2 → run (agent C)
             └── step 3 → agent A reads both results

The workflow is the coordinator, and it is deterministic. That keeps the hop count at one, and it makes the whole collaboration visible in the run tree.

Data model

TableHolds
agentsThe versioned agent spec: instructions, model, skills, tools, context policy, memory policy, budget.
skillsOne reusable procedure, its instructions, and the tools it needs.

Both are definitions. Neither holds execution state. Execution state is a run.

Risks

RiskMitigation
The tool list grows until it does not fit in a context windowThe registry serves a subset per agent, selected by the spec. A large surface is a selection problem, not a transport problem
A skill quietly becomes a prompt nobody reviewsA skill is a versioned row with an owner, and a change is a diff
An agent is given a plan it cannot execute, and it improvisesA plan names workflows and tools that exist. An unknown name fails at admission, not at step 7
Two agents grow two copies of the same ruleA rule is a policy row bound to an action. The agent spec carries budget defaults, never conditions
Model drift changes behaviour with no code changeThe model id is pinned on the spec, and a change is a new version

Open questions

  1. Who authors an agent? Engineers only, or any admin?
  2. Does a planner produce a new workflow, or only select one that exists? Generated workflows are powerful, and they are hard to review.
  3. Is a skill shared across agents by reference, or copied at version time?
  4. Can an agent propose a change to its own spec, for a person to approve? That is a useful loop, and it needs a hard policy boundary.

Diagram source. ac-docs/diagrams/mission-control-layers.drawio, page mission-control-intelligence. Edit that page, then run ./scripts/export-diagrams.sh mission-control-layers. Update this page and the diagram together.