Proposed specificationSource-located exemplarH1

Orchestrator-workers

Generate and delegate subtasks dynamically.

JOIN POINTdynamic_task_decompositionRECORD STATUSCandidate · v0.1.0

UML class view

The proposed contract drawn as participant roles, received calls, and numbered dependencies.

Object interactions

The same roles and calls as a class collaboration and a numbered message sequence.

Object interaction diagram for Orchestrator-workers, showing role collaboration and a numbered message sequence — FIG. W04 / OBJECT INTERACTIONS
Open full SVG Download PDF

Pointcut

Where the mechanism applies. All four dimensions belong to the proposed match, not just its vocabulary.

Context: A lead agent can generate tasks rather than call a fixed pipeline.
Trigger: The orchestrator decomposes a goal or dispatches a subtask.
Structure: Task plan, worker selection, delegation envelope, and synthesis step.
Effect: Synthesized output accounts for generated task results and delegated authority.

Protected contract

The obligation that clients, operators, and auditors should be able to rely on.

Requires: Workers advertise capabilities and share a task/result protocol.
Ensures: Every dispatched task has a result or failure and the final answer links them.
Invariant: Dynamic dispatch cannot create unbounded uncounted worker expansion.
Violation: Unsplittable, failed, or duplicate work is reflected in synthesis.

Negative control

A fixed sequence of sub-agents is W01 rather than dynamic W04.

Variation & lifecycle

The hot spot is admitted only where the protected semantics survive.

Permitted variation

Planner-executor, manager-worker, sub-agent registry, and dynamic graph.

Test obligation

Test decomposition quality, worker failure, duplicate task, and recursion cap.

Evolution question

Did fixed chains evolve into runtime task planning?

Prompt + code design

An illustrative prompt and a small design sketch for the proposed contract. Names and APIs are invented; this is not a repository excerpt.

PROMPT SKETCH

Decompose the task into worker-sized units, dispatch each with a task record, and require a result or failure for every dispatched task before composing the answer.

CODE DESIGN / PYTHON

def orchestrate(task, workers):
    tasks = planner.decompose(task)
    pending = {item.id: workers.dispatch(item) for item in tasks}
    results = {}
    for task_id, handle in pending.items():
        results[task_id] = handle.result_or_failure()
    if any(isinstance(result, Failure) for result in results.values()):
        return PartialResult(results)
    return composer.compose(results)

Illustrative design sketch. Dynamic decomposition is bounded by the result ledger, not by the planner's confidence. This example is a teaching aid, not implementation evidence or a conformance claim.

Detection & matching

The frozen rule that retrieves this record, the canonical source exemplar, and the control that should stay unmatched.

Join point: dynamic_task_decomposition
Source locator: Anthropic
Lexical cue: (?i)\b(?:orchestrator|orchestrate|task[_-]?decomposition|worker[_-]?agent|dynamic[_-]?dispatch)\bFrozen expression, v0.1.0. A match is a retrieval cue, not a confirmed implementation.
Aliases: None
Canonical exemplar: OpenHands/software-agent-sdkexamples/01_standalone_sdk/42_file_based_subagents.py:24Support: tests/sdk/subagent/test_subagent_registry.py:1
Negative control: A fixed sequence of sub-agents is W01 rather than dynamic W04.

Scope-anchored review. pohe-agent analyze <repo> --pattern W04 anchors the cue to its enclosing declaration and checks a separate operational signature for all four pointcut dimensions. Missing roles and negative-control risks remain visible; a structural match is not E2/E3. Calibration report ↗

Repository files

Up to five public files retrieved by the frozen cue and pinned to the audited commit. A match is a place to inspect, not a confirmed implementation.

01OpenHands/software-agent-sdkexamples/01_standalone_sdk/42_file_based_subagents.py:24CANONICAL SOURCE 02aaif-goose/goosecrates/goose-cli/src/cli.rs:1136RETRIEVAL CUE 03agno-agi/agnolibs/agno/agno/eval/accuracy.py:302RETRIEVAL CUE 04bytedance/deer-flowbackend/app/gateway/auth/oidc.py:429RETRIEVAL CUE 05china-qijizhifeng/agentic-harness-engineeringevolve.py:3715RETRIEVAL CUE

Evidence & provenance

Each layer answers a different question. Do not substitute retrieval for review.

Source-located exemplar

A concrete source anchor is available for this proposed record. It is one exemplar, not evidence that all repositories—or every contract field—conform.

Review boundary: this candidate and its specification come from an AI-assisted study. Independent human validation remains open.

ANY-EVIDENCE FAMILY COVERAGE30/36

83.3% of repository families in the frozen panel have at least one current cue, historical cue, or canonical source layer. This is not confirmed prevalence.

Current lexical-cue repositories: 33 / 41

Canonical source anchorOpenHands/software-agent-sdk 5c524ad65d81examples/01_standalone_sdk/42_file_based_subagents.py:24

Supporting surfacetests/sdk/subagent/test_subagent_registry.py:1

Primary agent delegates to registered subagents.

Exploratory retrieval cue Regex / v0.1.0

(?i)\b(?:orchestrator|orchestrate|task[_-]?decomposition|worker[_-]?agent|dynamic[_-]?dispatch)\b

This expression is a discovery instrument only. A match is a retrieval cue—not a confirmed implementation, pointcut match, or contract proof.