Skip to content
DarkPrint
Aautogentask-decomposer1.0.0

Task Decomposer

Get card

Clone

npx -y darkprint clone task-decomposer@1.0.0

There is no repository and no history behind a card. Download hands you the document as it stands, and Clone fetches the same document by name.

↓ 1 downloads

Break the request into an ordered set of sub-tasks small enough for one solver pass each, and carry the acceptance criteria down onto every one of them.

used in 1 blueprint

Specification

119 words · handed to the agent

Read the stamped brief on request and split its goal into an ordered list of sub-tasks, each small enough that one pass can finish it and each stating its own done condition. Emit at most six; when the goal will not fit in six, merge the smallest adjacent steps instead of emitting a seventh. Onto every sub-task copy the slice of the run's criteria reference that sub-task answers to, so each one can be judged on its own without re-reading the whole brief. Emit the ordered list on subtasks, and decompose the same brief the same way every time: two solvers work this plan in parallel, and their answers are only comparable if they started from an identical plan.

Interfaces

1 in · 1 out

Inputs

1
Inputs declared by this node card
NameData typeRequiredDescription
requestjson requiredThe stamped task brief, straight from intake.

Outputs

1
Outputs declared by this node card
NameData typeDescription
subtasksplanThe ordered sub-tasks, each carrying the slice of the acceptance criteria it answers to.

Dependencies

1

The upstream nodes this card expects to receive from. Whenever a blueprint pins this card, each name is checked against a real edge in that graph. A name without a link is not a published card; it refers to a node inside some graph.

Card values

17 declared

Who the node is. The id is the key the DOT pins.

id
task-decomposer
name
Task Decomposer
type
agent
phases
Planning
Behaviourcard spec →

What it does, and the prose the agent is handed when the graph runs.

action27 words
Break the request into an ordered set of sub-tasks small enough for one solver pass each, and carry the acceptance criteria down onto every one of them.
spec119 words
Read the stamped brief on request and split its goal into an ordered list of sub-tasks, each small enough that one pass can finish it and each stating its own done condition. Emit at most six; when the goal will not fit in six, merge the smallest adjacent steps instead of emitting a seventh. Onto every sub-task copy the slice of the run's criteria reference that sub-task answers to, so each one can be judged on its own without re-reading the whole brief. Emit the ordered list on subtasks, and decompose the same brief the same way every time: two solvers work this plan in parallel, and their answers are only comparable if they started from an identical plan.in full above
model
claude-opus-5
agent
Planner
skill
skills/task-decomposer.md
tools
none
mcp
none
params
temperature: 0.2, max_subtasks: 6
Interfacescard spec →

What arrives, what leaves, which nodes it expects to hear from, and what may not.

inputs
request : json
outputs
subtasks : plan
dependencies
task-intake
cannot
no type is refused
will_not
decompose the same brief two different ways
Evaluation metadatacard spec →

The keys the static analysis reads. Nothing here instructs the agent.

risk_markers
none
notes24 words
Low temperature on purpose. The adversarial spread belongs downstream between the two solvers; a plan that differs run to run makes their proposals incomparable.
Service fieldscard spec →

The card's own version, and who wrote it.

version
1.0.0
author
autogen
provenance
not stated

Definitions for every card field

Version history

1 version published
  1. task-decomposer@1.0.0currentsha256:0323d2baae0c689df113561b6a104a794c056ea749980f351e20756a84aa91d1

    pinned byAdversarial Consensus Lineautogen/adversarial-consensus-line

A digest is a fingerprint (SHA-256) of the card's content, computed without the author and provenance fields. The same card from two people gets the same digest; any edit gets a new one.

First published version, so there is nothing to compare yet. Versions are never edited in place: the next change arrives as a new version, and the differences between the two documents are listed here.

Community notes (0)

No notes yet.

Nobody has posted about this node card yet.

Sign in to post a note.