Skip to content
DarkPrint
Aautogenconfidence-escalation1.0.0

Confidence Escalation

Get card

Clone

npx -y darkprint clone confidence-escalation@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

Let anything at or above the confidence threshold through untouched and park the rest for a person, with the full context pack attached and an SLA on the clock.

used in 2 blueprints

Specification

93 words · handed to the agent

Read the confidence figure on the verdict you are handed. At 0.7 or above, pass the work through on accepted exactly as it reached you and take no other action. Below 0.7, decide nothing yourself: assemble the parked case on escalated with the work that was graded attached, taken from the context you were given, or pulled from the run record when none was wired, put a fifteen-minute deadline on it, notify the reviewer on call, and wait. Nothing reaches accepted that has not either cleared 0.7 or been approved by that person.

Interfaces

2 in · 2 out

Inputs

2
Inputs declared by this node card
NameData typeRequiredDescription
verdictstatus requiredThe upstream check's confidence signal, the number this node reads.
contextjson optionalThe work that was graded, attached to the escalation so nobody has to go digging.

Outputs

2
Outputs declared by this node card
NameData typeDescription
escalatedjsonThe parked case as handed to a human, with its context pack and deadline.
acceptedjsonThe work passed straight through, untouched, because confidence cleared the bar.

Dependencies

1
  • qa

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

15 declared

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

id
confidence-escalation
name
Confidence Escalation
type
human-gate
phases
none declared
Behaviourcard spec →

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

action29 words
Let anything at or above the confidence threshold through untouched and park the rest for a person, with the full context pack attached and an SLA on the clock.
spec93 words
Read the confidence figure on the verdict you are handed. At 0.7 or above, pass the work through on accepted exactly as it reached you and take no other action. Below 0.7, decide nothing yourself: assemble the parked case on escalated with the work that was graded attached, taken from the context you were given, or pulled from the run record when none was wired, put a fifteen-minute deadline on it, notify the reviewer on call, and wait. Nothing reaches accepted that has not either cleared 0.7 or been approved by that person.in full above
model
whatever the graph supplies
agent
not named
skill
not named
tools
Human review, Messaging
mcp
slack
params
sla_minutes: 15, confidence_threshold: 0.7
Interfacescard spec →

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

inputs
verdict : status, context : json
outputs
escalated : json, accepted : json
dependencies
qa
cannot
no type is refused
will_not
decide a parked case itself
Evaluation metadatacard spec →

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

risk_markers
none
notes49 words
context is optional because an edge carries a single value: the QA step wires its verdict, and the escalation pulls the draft it graded from the run record. The human-gate type is the whole of what says a person acts here, and nothing on the card can contradict it.
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. confidence-escalation@1.0.0currentsha256:96cbd4b93d6dfc77beea261e7ecdbad8967ffe02df05641d105f31be4183623a

    pinned byFrontline Triageautogen/frontline-triage,Incident Commanderautogen/incident-commander

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.