Ayodhya Mohanthy product designerInstall skills ↗

01 · HARNESSES & WORKFLOWS TESTED PRINCIPLES

Design the environment
around the model.

The product is the loop: context, tools, handoffs, memory, checkpoints, and the interface that makes each step legible.
IntentPlanContextTool callObservationState updateCheckpointContinue / Ask / Stop

HUMAN CONTROL · STEER · APPROVE · INSPECT · RECOVER

WHY THIS CHANGES THE PRODUCT

A polished chat surface cannot repair a weak loop.

Context should be selected for the next decision, not accumulated without limit. Tool interfaces are product interfaces: names, parameters, responses, and errors shape agent behavior. Handoffs must carry the request, state, evidence, restrictions, and unresolved decision, not merely a summary.

Memory, progress, and permissions are separate system concerns. If the interface collapses them into a confident answer, it hides the product's real state.

FIVE TOPICS

01

The loop is the product

What happens between intent and completion?

02

Context is a decision system

What must the agent know now?

03

Tools are contracts

Can the agent call and recover without guessing?

04

Handoffs preserve ownership

What survives when work moves?

05

Memory is not a transcript

What should persist, expire, or be rechecked?

PATTERNS TO USE

Observable agent loop

Every step has an owner, state, and stop condition.

RELEASE CHECK

Context budget

Freshness, relevance, and disclosure limits stay visible.

RELEASE CHECK

Tool contract

Inputs, outputs, errors, authority, and effects are explicit.

RELEASE CHECK

Handoff packet

Goal, owner, state, evidence, restrictions, and acceptance travel together.

RELEASE CHECK

Resume after interruption

Read current state before repeating an external effect.

RELEASE CHECK

Ask for the missing decision

Research first; ask only where a person must choose.

RELEASE CHECK

Source-of-truth refresh

Recheck action-driving facts at the system that owns them.

RELEASE CHECK
“Do not design the agent's answer. Design the conditions under which it can keep working without losing truth, control, or the next decision.”Ayodhya's take

FAILURE MODES

More context, less relevanceStale or contradictory action
Tool name hides effectWrong action or unsafe approval
Summary-only handoffLost restrictions and evidence
Progress inferred from narrationClaims work that did not land
Retry without idempotencyDuplicate external effect
Memory treated as authorityHabit becomes unauthorized action

PRIMARY SOURCES · REVIEWED 16 SEP 2026

AnthropicEffective context engineering for AI agentsAnthropicWriting effective tools for agentsOpenAIHarness engineeringOpenAIMCP and ConnectorsOpenAIAgents SDK handoffs
Ayodhya Mohanthy, product designer.An open reference for designing agent systems.Reference ↑