R079 · AI education › Teaching materials and glossaries

15 agent design patterns, checked for teaching

Read the original

teachingagent-architecturemulti-agentagent-workflowshuman-oversightlearning-resources

A useful reading list with a weak taxonomy. Maryam Miradi's post (LinkedIn, 2026-11-11) names 15 agent patterns in three tiers and links a paper for each. The papers are the value. The tiers blur several real distinctions: five of the fifteen are the same generate-critique-retry loop, and three are the same delegation tree. For teaching, regroup them by who owns control flow and what grounds the check.

The 15 patterns as posted

#Pattern (post's name)Linked sourceFit of the citation
1ReActReAct, Yao et al. 2022Exact
2Plan-and-ExecutePlan-and-Act, 2025Good: separate planner and executor models, with re-planning
3Reflection / self-critiqueReflexion, Shinn et al. 2023Good, but Reflexion learns from a task signal, not pure self-review
4Tool use / function callingGorilla, 2023Narrow: fine-tuning for accurate API calls
5Orchestrator-subagentLLM multi-agent survey, 2024Loose: a survey, not the pattern
6SupervisorAutoGen, 2023Loose: AutoGen is conversable agents; supervisor is one setup
7Hierarchical agentsAgentOrchestra (TEA protocol), 2025Fair: a central planner over sub-agents and versioned tools
8Sequential pipelineLangGraph docsA framework, not evidence
9Parallel fan-out / fan-inLLMCompiler, 2023Good: a DAG of calls run concurrently
10MapReduceSame survey as 5Reused link; same pattern as 9
11Debate / adversarialMultiagent Debate, Du et al. 2023Exact
12Evaluator-optimizerSelf-Refine, 2023Mismatch: Self-Refine is one model, not a separate evaluator
13Critic-actorCRITIC, 2023Good: critique grounded in tools
14Self-healing / retrySame as 3 (Reflexion)Reused link
15Human in the loopPlan-Then-Execute user study, CHI 2025Good: editing the plan before it runs changes trust and team outcomes

The post itself adds one-line definitions and a closing line per tier. It ends with a course promotion and a lead-magnet page. The "400+ agents built" claim is the author's and is not checkable.

What the catalog gets wrong

Missing patterns

Choosing a pattern

  1. Start with one model call plus retrieval and tools. Most tasks stop here.

  2. If the steps are known, write a workflow: a chain or a router, with code checks between steps.

  3. If the steps depend on what the model finds, use a single-agent loop (ReAct) with a step and token budget.

  4. Add a check loop only where an outside signal exists: tests, a schema, a tool, a rubric another model can score reliably, or a person.

  5. Split into sub-agents only for context isolation or parallel breadth, and price it. Multi-agent runs cost several times the tokens of one agent. The baseline to beat is best-of-N from one agent at the same budget (R019).

  6. Put a human at the step where a mistake is expensive or cannot be undone, typically the plan, not every action (the HITL study linked as pattern 15).

Failure modes by tier

TierFailureMitigation
Single agentLoops without progress; a wrong early thought steers later tool callsStep and token budgets; progress checks; re-plan on tool error
Single agentPlan goes stale as the world changesRe-plan after each observation that contradicts the plan
Multi-agentErrors compound down a pipeline; each hop drops contextTyped handoff artifacts; validate at each boundary
Multi-agentWorkers duplicate or contradict each otherDisjoint task specs; one owner merges results
Multi-agentDebate converges on a confident wrong majorityDiverse models or prompts; a judge with ground truth where possible
Feedback loopCritic and actor share blind spots (same model)Ground the critic in tools or tests; use a different model
Feedback loopLoop optimises the judge, not the task (R039, R066)Hold-out checks; cap iterations; measure regressions, not only gains
Human in the loopRubber-stamping after many approvalsFewer, higher-stakes checkpoints; show diffs and evidence

Review checklist

Teaching session plan (60 minutes)

  1. 10 min, regroup. Hand out the 15 names. Ask the group to merge duplicates. Land on four families: workflows (chain, router, fan-out), the agent loop (ReAct, plan-and-execute), delegation (orchestrator, supervisor, hierarchy, handoff) and check loops (reflection to debate, plus HITL).

  2. 15 min, one loop, four groundings. Same code-generation task with self-review, a second model, unit tests and a human. Compare which catch the planted bug.

  3. 15 min, workflow vs agent. Build the same task as a three-step chain and as a ReAct agent. Compare tokens, latency and how easy a failure is to trace.

  4. 10 min, when to split. Show a task where sub-agents help (wide parallel research) and one where they hurt (a tightly coupled edit).

  5. 10 min, critique the post. Use the citation-fit column above: which links support their pattern, and which are a survey or a framework doc standing in for evidence.

Use the post as a list of names and the linked papers as reading. Link to the post rather than copying its infographic.

Sources

Connections

See R079 on the knowledge map

Part of the Research notes on nowucca.com. Figures and claims are the source's own unless the summary says otherwise.