Full report

The full report is the complete record of one assessment run: the overall risk, the attack surface that was tested, and every finding with its evidence, recommended fix, and standards mapping.

Open the full report

From a run's details page, select Open full report. See Run details.

Figure 1. Select Open full report.

The report has three sections, from top to bottom: security posture, capabilities exercised, and findings.

Security posture

The top of the report gives the overall risk evaluation for the run.

• Risk grade and score: the letter grade in the ring, with the risk score below it, such as C and risk 17/100. A lower score means less risk.
• Status: the run's status and outcome, such as Complete · Residual risk, meaning the run finished and some risk remains.
• Target and engine: the asset that was tested and the red-team engine that tested it, such as the Quilr Adaptive Engine in adaptive mode.
• Result counts: how the tested objectives ended.

Count	Meaning
Breached	The attack succeeded. These are confirmed risks.
Partial	The attack partly succeeded.
Held	The target withstood the attack.
Inconclusive	The result could not be decided.
Tested	The total number of objectives tested.

To download the report, select PDF or Markdown at the top right.

Figure 2. Security posture.

Capabilities exercised

This section shows the attack surface the assessment actually tested, and which techniques from each standard were used.

At the top, a chip shows the input channel tested, such as Text. For a text-only target, the report notes that you can enable multimodal objectives to also test vision, audio, and document channels.

Below it, six counts summarize the scope of the run:

Count	What it shows
Objectives run	Attack objectives the run executed.
Tools probed	Tools the run tried to make the target misuse. This is 0 when no tools were configured.
OWASP categories	OWASP LLM and agentic categories covered.
ATLAS techniques	MITRE ATLAS techniques used.
NIST AI RMF cited	NIST AI RMF subcategories referenced in the results.
ATLAS mitigations	MITRE ATLAS mitigations recommended in the findings.

The panels below break the run down further:

• Severity breakdown: findings by severity: Critical, High, Medium, and Low.
• OWASP LLM & Agentic coverage: each OWASP category tested. A red dot marks a category where an attack succeeded; green marks one the target held.
• MITRE ATLAS coverage: each ATLAS technique used, with the same red and green markers.
• Tool exposure: each configured tool. A tool shown in red is one the agent was induced to call during the attacks.
• Attack-technique effectiveness: which adversarial techniques succeeded, and how many times. Prioritize your defenses against the techniques at the top.

Figure 3. Capabilities exercised.

Findings

The last section lists every issue the run surfaced.

• Use Search findings to find a finding by name, severity, or risk category.
• Select All findings, or Breached & partial to show only attacks that fully or partly succeeded.

Each finding shows its title and severity, then the OWASP category and MITRE ATLAS technique it maps to, the technique that succeeded (such as via Direct), and whether reproduction was attempted. A Synthesized tag marks a finding from an objective generated specifically for this target.

Figure 4. The findings list.

Inside a finding

Select a finding to expand it.

Part	What it shows
Evaluation provenance	How the result was judged: the evaluation mode (such as hybrid), the number of hard signals detected, the judge consensus, and whether the judges disagreed.
Reproduction	Whether Quilr tried to reproduce the attack. Reproduction is not attempted on a live target unless it is declared a replay-safe sandbox.
What happened	A plain-language account of the attack and how the target responded.
Evidence	The exact part of the conversation that proves the finding, with its turn number.
Recommended fix	What to change so the target no longer behaves this way.
OWASP mitigation	The mitigation for the mapped OWASP category.
MITRE ATLAS mitigations	The ATLAS mitigations for the mapped technique, with their IDs, such as AML.M0002.
NIST AI RMF	The NIST AI RMF subcategories that apply, such as MEASURE 2.7.

Figure 5. An expanded finding.

tip

Share the recommended fix and evidence with the development team that owns the target. After they apply the fix, replay the finding from the run's details page. See Run details.