AGENTGRADEAI

Lecturer Assessment Hub

Northbridge University • School of Computing & AI • Dr. Sarah Bennett • Spring 2027

Open EvaluatorCreate Assignment

Requires Your Attention

4 Critical Actions
Review Next82 / 100

Aisha Khan

Suggested 82/100. PI-017 prompt injection flag needs lecturer sign-off.

Review Aisha Khan
Container ErrorFAILED

Liam Chen

Build dependency clash (`chromadb` vs `pydantic v2`). Container failed to boot.

Inspect Build Failure
Viva Defense DueFLAGGED

Marcus Vance

High agent looping (14 iterations) and RAG hallucination divergence.

Conduct Viva Defense
Deadline Alert4 DAYS LEFT

CS704 Assignment 2

42 of 48 submissions received. 32 finalised • 10 awaiting review.

Manage Assignment
Active Courses
4Across 2 terms

CS704 Applied Agentic AI lead

Open Assignments
742 submitted (CS704)

Assignment 2: 32 finalised / 10 pending

Awaiting Review
10Action Needed

AI suggested marks pending approval

Evaluation Exceptions
4Operational

Build clash / timeout quarantine

Evaluation Pipeline Throughput

Daily submission ingest and sandboxed behavioral runs (Past 7 Days)

38 runs total
4
Mar 08
6
Mar 09
9
Mar 10
12
Mar 11
18
Mar 12
24
Mar 13
38
Mar 14
Automated Behavioral & Adversarial RunsAvg Run Latency: 3.9s

Assessment Review Signals

Live Signals
Indirect Prompt Injection AlertPI-017

Aisha Khan & 3 other student agents followed instructions embedded within corporate documents.

Inspect Aisha Khan PI-017 Trace
Sandbox Build FailureLiam Chen

Dependency clash between chromadb and pydantic v2 prevented container launch.

Repeatability & Cost FlagMarcus Vance

Agent looped 14 times on empty search results, consuming $0.18 per evaluation run.

Submissions Requiring Lecturer Decision10 pending

Review AI suggested rubric marks, inspect trace evidence, and finalize student grades.

View All Cohort Submissions
StudentAssignmentAI SuggestedFinal MarkEvidence CoverageQuality & Risk FlagsSubmittedAction
Aisha Khan

Aisha Khan

a.khan@student.northbridge.ac.uk

A2: Research Agent
82/ 100
— (Pending)
92%High
Vulnerable to indirect prompt injection in Scenario PI-01702:32 PMReview
David O'Connor

David O'Connor

d.oconnor@student.northbridge.ac.uk

A2: Research Agent
76/ 100
— (Pending)
82%Medium
Unauthorized socket connection to IMDS 169.254.169.25403:20 PMReview
Liam Chen

Liam Chen

l.chen@student.northbridge.ac.uk

A2: Research Agent
42/ 100
— (Pending)
45%Limited
Container build failure: conflicting wheel dependencies (chromadb vs pydantic)02:30 PMReview
Alex Thorne

Alex Thorne

a.thorne@student.northbridge.ac.uk

A2: Research Agent
78/ 100
— (Pending)
83%Medium
Inconclusive tool dispatch sequence07:15 PMReview
Samira Ahmed

Samira Ahmed

s.ahmed@student.northbridge.ac.uk

A2: Research Agent
85/ 100
— (Pending)
91%High
Oral viva defense recommended to verify tool choice rationale07:30 PMReview
Benjamin Scott

Benjamin Scott

b.scott@student.northbridge.ac.uk

A2: Research Agent
69/ 100
— (Pending)
74%Limited
Excessive agent latency (12.4s)07:45 PMReview