Vulnerability scan report summaries and fix lists
An agent reads long vulnerability scan reports, groups findings by system and severity, removes repeats, and drafts a short fix list in plain English for the person who maintains those systems.
The use case
Small businesses that run vulnerability scans, often because a customer or insurer requires it, receive long reports full of repeated findings and technical language. The report gets filed and little gets fixed. The agent reads each new scan export, groups findings by system, removes duplicates across hosts, and compares the results with the previous scan to show what is new, what was fixed, and what is still open. For each group, it writes a short plain-English description and a likely fix, such as applying a specific update or disabling an old protocol. It orders the list using the scanner's severity and whether the system faces the internet. The IT provider or system owner decides what to fix and when, and makes all changes. The agent does not change systems or accept risk on anyone's behalf.
Example workflow
- After each scheduled scan, the report is exported from the scanner as CSV or JSON.
- The agent groups findings by system and vulnerability, removing duplicates across hosts.
- It compares the results with the previous scan to mark findings as new, fixed, or still open.
- For each group, it writes a plain-English summary and likely fix, ordered by severity and internet exposure.
- The fix list is posted to the IT ticket queue, and the system owner reviews, schedules, and completes the work.
- Findings the owner decides not to fix are recorded with a reason and review date by the owner.
Tools
- Tenable Nessus
- Qualys
- Jira
- Google Sheets
- Claude
- Slack
Risks and guardrails
- A suggested fix is wrong for the specific system — guardrail: each item links to the scanner's original finding and references, and the system owner decides the fix.
- Severity is understated, so an urgent issue waits — guardrail: the agent never lowers the scanner's severity rating; it can only add context.
- Scan reports describing weaknesses are shared too widely — guardrail: reports and fix lists are stored in a restricted folder and ticket project.
- Accepted risks are recorded without an owner — guardrail: only a named person can mark a finding as accepted, with a review date.
Related recipe
Security Questionnaire First-Draft Agent with Cited Sources: An agent drafts security questionnaire answers from your own trust documents and approved answer library, cites every source, and marks gaps NEEDS OWNER for a responsible person to sign.
Want it set up for you? Request a quote.
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- Claude (official page): claude.ai (checked 2026-11-09)
- Slack MCP server (official page): docs.slack.dev/ai/slack-mcp-server (checked 2026-11-09)