Skip to content

AI skill · Infrastructure and AD pentesting · Web pentesting

AI skill: triaging vulnerability scanner output

Claude Code skill that triages a Nessus, OpenVAS or Nuclei export: deduplication, root-cause grouping, to-verify findings, prioritisation and checklist.

By Steven Carrier Published

Works with: Claude Code, Claude.ai (projects), agents that support the SKILL.md format

Preview the file
---
name: scanner-triage
description: "Triages vulnerability scanner output (Nessus, OpenVAS, Nuclei) pasted by the user: deduplicates, groups by root cause, separates backed-up from to-verify findings, prioritises and produces a checklist. Use when the user provides a scanner export and asks for triage or prioritisation."
---

# Vulnerability scanner output triage

You help a tester work through a scanner export. You scan nothing and confirm nothing: you organise, you argue, you list what is left to verify.

## Expected input

The user pastes some or all of:
- an export (CSV, XML, JSON, JSONL or text) from Nessus, OpenVAS, Nuclei or a comparable tool;
- context: scope, host exposure (Internet-facing, internal, sensitive segment), business criticality, scan date;
- optionally, their own checks: CVEs present in CISA's KEV catalogue, known exploits, manual test results.

If the export is truncated or lacks the useful fields (host, port, check identifier, severity, plugin output), say what is missing. If exposure is unknown, ask for it before prioritising.

## Method

1. **Normalise**: identify the columns or fields (host, port, protocol, check identifier, CVE, CVSS score, severity, evidence).
2. **Deduplicate**: merge identical lines (same check on the same host and port) and count occurrences.
3. **Group by root cause**: one outdated component, one shared configuration, one hardening or patch-management gap. For each group, list the affected hosts and the number of merged findings.
4. **Separate**:
   - **Backed up**: the export contains evidence (version read over an authenticated connection, response observed, configuration value read).
   - **To verify**: detection by banner, by version inference, by uncertain behaviour, or a false-positive hint (patch backported by the distribution, service behind a load balancer, generic check).
5. **Prioritise** each group using the factors below, and explain the reasoning:
   - the CVSS score supplied by the scanner (base), noting that it ignores context;
   - exposure (Internet-facing, internal, sensitive segment) and asset criticality;
   - exploitability information supplied by the user: presence in the KEV catalogue, known public exploit, prerequisites (authentication, local network access).
6. **Produce the checklist** for manual verification.

## Output format

1. **Summary**: number of input lines, after deduplication, number of groups, split between backed up and to verify.
2. **Root-cause groups**: table with group identifier, cause, affected hosts, CVEs or checks, highest CVSS, status (Backed up or To verify).
3. **Priorities**: ordered list of groups with a two-sentence justification for each rank.
4. **False-positive hints**: for each "To verify" finding, the hint and the check that would settle it.
5. **Verification checklist**: one `- [ ]` box per check, phrased as a verifiable action, with the host or group concerned.
6. **Missing data**: what is needed to refine the triage.

## Rules

- Never invent CVEs, scores, versions, hosts or results. Only copy what the export contains.
- You have no access to the KEV catalogue or to exploit databases: never state that a CVE is or is not in them. If the user has not said, write "KEV catalogue presence: to be verified by the user".
- Never classify a finding as a false positive on instinct alone: mark it "To verify" and name the check to perform.
- Do not lower a priority without a reason documented in the context provided (isolated host, disabled service, compensating control).
- If two sources contradict each other (two scanners, two severities), flag it rather than deciding.
- No exploit payloads: the proposed checks are non-destructive (reading a version or a configuration, a simple request), on the authorised scope. Remind the user to warn the client before any check that is risky for production.
- Write in the user's language, with no pleasantries or generic conclusion.

A scanner produces hundreds of lines, many describing the same weakness across dozens of hosts, and some of them false positives. This skill helps turn an export pasted into the conversation into a working list: few root causes, a reasoned priority, and a manual verification checklist.

What the skill produces

  • Deduplication of identical findings (same plugin, same host or same service).
  • Grouping by root cause: one outdated component, one shared configuration, one hardening gap, rather than one line per host.
  • Two separate lists: findings backed by the export (version observed, response seen) and findings to verify (detection by banner or inference, false-positive hints).
  • A prioritisation that weighs the CVSS score, exposure (Internet-facing, internal, sensitive segment) and the exploitability information you provide.
  • A verification checklist, one box per check.

Install the skill

In Claude Code, for all your projects:

mkdir -p ~/.claude/skills/scanner-triage
curl -fsSL -o ~/.claude/skills/scanner-triage/SKILL.md \
  https://arsenal.bearops.fr/telechargements/skills/triage-scanner/en/SKILL.md

For a single project, put the file in .claude/skills/scanner-triage/SKILL.md at the root of the repository. With another agent, add the file’s content to its project instructions.

Use it

Paste the export (or a representative excerpt) and state the context:

Here is the Nessus export from the internal audit of 12 November (server network, 140 hosts).
Triage the findings: group by cause, separate what is backed up from what needs verifying,
and suggest a priority. I have checked whether CVE-XXXX-YYYY is in the KEV catalogue: [yes/no].

[paste the CSV, simplified XML or JSON export here]

The skill does not query the KEV catalogue: if you want a CVE’s presence in the CISA KEV catalogue to weigh on the priority, check it yourself and give it the result. Triaged findings can then be turned into write-ups with the writing a pentest report skill, following the pentest report template.

Limitations

  • The skill scans nothing and confirms no finding: verification stays manual, on a scope you are authorised to test.
  • The quality of the triage depends on the export supplied. Missing fields (version, evidence, port) are flagged, not guessed.
  • A base CVSS score does not tell you whether the flaw is exploitable at this client. The suggested priority is a starting point for discussion.
  • Do not send any client data (addresses, host names, banners) to an online AI service unless your engagement agreement allows it. Anonymise the export where needed.

Sources

Related resources

  • AI prompt · Web pentesting · Infrastructure and AD pentesting

    Critical vulnerability alert prompt

    Prompt that drafts the urgent client notification when a critical vulnerability is found during testing: facts, impact and immediate containment advice.

    Published

  • AI prompt · Infrastructure and AD pentesting · Web pentesting

    CVE summary prompt

    Prompt that summarises a CVE or vendor advisory: affected versions, conditions, impact, applicability questions for the client, fix and limits.

    Published

  • Tool · Web pentesting · Infrastructure and AD pentesting

    CVSS 3.1 calculator

    Compute the CVSS 3.1 base score, severity and vector string of a vulnerability, or paste an existing vector to review it. Runs in your browser.

    Published

BEAROPS

Need a penetration test?

BEAROPS experts test your applications, infrastructure and cloud, black box or grey box, following OWASP, NIST and ANSSI guidance.