Capabilities / D · Medical practice

D.17 Red-flag recognition and triage Proposed

Given case material or the record of a completed consultation, the agent fills a structured triage form stating urgency, disposition and whether emergency services must be called; emergencies must be assigned the highest level. Graded by checkpoints on each triage field.

Proposed. This capability is under RFC and does not yet carry levels.

Evidence: none. No benchmark card lists this capability yet. No level is issued: methodology 0.1.0 sets no level thresholds.

Domain
D · Medical practice
Evidence form
checkpoint
Veto constraints
X.9 Medical safety boundary
Depends on
A.3 Discrepancy detectionA.4 Multi-source cross-analysis
Required by
—
Internal task families
—
External anchors
HealthBench
Version
0.1.0 · 0.1.0

Definition in detail

Under-triage of an emergency is the most consequential error a medical agent can make. Triage must be stated in structured fields so that it can be checked, not inferred from prose.

Inputs a task gives

  • Case material such as history, vital signs and examination findings
  • The record of a completed consultation, supplied as a static document
  • Triage level definitions and disposition options

Artifact required

A triage form: urgency_level, disposition, call_emergency_services (yes or no), red_flags[] (finding, evidence as a verbatim excerpt) and the rule each red flag triggers.

How it is graded

Checkpoints compare each triage field with the reference; where a task cites grading references, there is one checkpoint per reference. An emergency assigned anything below the highest level fails the item. X.9 is a one-shot veto.

In scope

  • Recognising red-flag findings in the supplied material
  • Assigning urgency and disposition
  • Deciding whether emergency services must be called

Out of scope

  • Listing and ranking differential diagnoses - graded under D.18
  • Disclosure items such as indications for return - graded under D.20
  • Responses to out-of-scope or dangerous requests - graded under X.9

Typical failures

Expected failure patterns; not measured rates.

Standards: std-aha-acc-2021-chest-pain · std-cn-chest-pain-consensus-2019

Benchmarks

No benchmark card lists this capability.

Grading references

No grading reference is mapped to this capability.

Levels by model and harness

No verified results yetLevels would be derived from site-verified result records; methodology 0.1.0 sets no level thresholds, so the level shows not issued.