Works with the AI Agents you already use
CRA Risk Analyzer
CRA Risk Analyzer brings vulnerability (CVE) risk analysis into **Claude Code**, **Codex** and **GitHub Copilot** (agent mode).
Free
CRA Risk Analyzer
Example session with this skill installed
Assess the risk of CVE-2023-3519 on host web-prod-01 and generate a risk report.
- Read your context and instructions
- Compiled the cra risk analyzer
- Generated the data export
I have completed the assessment for CVE-2023-3519 on host web-prod-01. The contextual score has been calculated, and the risk reports have been generated.
Risk Assessment: CVE-2023-3519 on web-prod-01
| Metric | Value | Note |
|---|---|---|
| CVE ID | CVE-2023-3519 | Citrix ADC RCE |
| Host Slug | web-prod-01 | (Pseudonymized in report) |
| CVSS Baseline | 9.8 | CRITICAL |
| Contextual Score | 8.2 | HIGH (Mitigation: WAF active) |
| EXPOSED-RCE | YES | Remote Code Execution possible |
| Prob. (Year) | 0.142 | 14.2% annual probability |
| Expected Loss | 12,450.00 | Currency: USD |
| Remediation Due | 2024-05-15 | Status: ON-TIME |
Attack Chain Summary
The exploit requires an unauthenticated network request to the management interface. Chaining with CVE-2023-24489 is possible if the local user profile is accessible, increasing the probability of full system compromise.
Files Created
assessments/2024-05-01-cve-2023-3519-web-prod-01.mdreports/2024-05-01-risk-summary.mdreports/2024-05-01-risk-summary.csv
Next steps
- Review the
mitigation-guidancesection in the CVE card. - Apply the recommended firmware update before the remediation due date.
- Run
cra-reportagain after patching to verify risk reduction.
2024-05-01-risk-summary.csv
CSV · data export
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
CRA Risk Analyzer brings vulnerability (CVE) risk analysis into Claude Code, Codex and GitHub Copilot (agent mode). Seven Markdown skills: no application to host, no bundled scripts, nothing to install beyond copying the skill folders. Just ask — the cra entry point recognises which of three questions you are asking:
- One CVE — "tell me about CVE-2021-44228": a short answer with the attack vectors, mitigations and fixed versions, an exploitability verdict with its reasons (network, no authentication, known exploited, public exploit…), a reference risk without host context, and the classes of CVE it can be chained with. No setup questions: it starts at once with defaults.
- A CVE list for one host — paste the CVE IDs from a report and name the host: risk, mitigation and an exploitable / conditional / not-exploitable-here verdict for each CVE, plus the host's combined risk and the chains between its own CVEs.
- A full scanner report — Nessus, OpenVAS, Qualys, Trivy, Grype, SARIF, CycloneDX, CSV or JSON: a triage summary first (severities, distinct CVEs, hosts, known-exploited matches), then a full analysis of the High+ and known-exploited CVEs, up to 50 per run, with the rest listed as not analysed.
What you get
- cra: the entry point. Recognises the three modes, routes your request, shows the catalog and holds the shared rules.
- cra-setup: configures the workspace: workspace folder, risk option (score, probability or money), remediation windows per severity, threat context, currency and asset values, privacy. A single-CVE question needs none of it; the questions come before your first scan import or host assessment.
- cra-cve: a sourced card for one CVE from NVD, CISA KEV and FIRST EPSS: attack vectors, mitigations, the upstream versions that fix it, an exploitability verdict with reason codes and a host-less reference risk.
- cra-vector: maps a CVE to attack steps and finds chains with other CVEs in your own estate; with nothing to pair it with, it names the classes of CVE that would complete a chain (never an invented CVE ID).
- cra-scan: imports a scan report (CSV, JSON, or native reports from Nessus, OpenVAS/Greenbone, Qualys, Trivy, Grype, SARIF and CycloneDX/VEX, detected by content), groups CVEs per host with each finding's component and versions, and writes a triage summary for multi-host or large reports.
- cra-assess: takes an imported scan or a pasted CVE list for one host, asks a short mitigation survey (with a one-answer bulk option for large selections), then gives a deterministic contextual score (flagging internet-facing, unpatched remote code execution as EXPOSED-RCE), a severity with a remediation due date (escalated for known exploited CVEs, flagged OVERDUE once passed), probabilities, a risk factor and an expected loss, a fix verdict with the minimum fix version, an exploitability verdict per CVE, and a host summary with the host's combined risk and its chains.
- cra-report: a Markdown report (risk table with
FixandExploitablecolumns, hosts with their combined risk and chains, the CVEs not analysed) and a CSV with a fixed column order.
How it works
- Ask:
cradecides the mode — one CVE, one host's list, or a full report — and runs the matching path. Without a workspace, a single-CVE question starts at once with defaults. - Set up (first scan or host assessment):
cra-setupdetects what your environment can do, asks only what it cannot detect, and writescra/environment.mdplus an editablecra/risk-parameters.md. - Look up:
cra-cvewrites one card per CVE with CVSS, KEV listing, EPSS, fixes and fixed versions, an exploitability verdict and a reference risk, each fact with its source and fetch time. - Map:
cra-vectoradds attack steps and the chains a CVE forms with others on the same host or in your estate. - Import:
cra-scanturns a scanner export into a scan file grouped per host, with a triage summary for large or multi-host reports — or paste a CVE list for one host andcra-assessbuilds the scan file itself. - Assess:
cra-assessasks a few mitigation questions, shows whether the installed version already carries the fix (your answer still decides), then computes the score, severity, due date, exploitability, probabilities, expected loss and each host's combined risk. - Report:
cra-reportwrites the Markdown report and the CSV.
Every number is a Markdown or CSV file in your project, with its inputs and formula next to it, so you can read, diff and recompute each one by hand. The contextual score never exceeds the CVSS baseline, unknown controls never lower the risk, and probability parameters come from a documented rubric (KEV, exploit evidence, EPSS, your answers), not from guesses.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.
Requires local filesystem for workspace files. Markdown/CSV only for reports. NVD_API_KEY recommended for rate limits. Does not search the web for external exploit chains.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 9 days ago
- Free with an Agensi account
- Accounts
First
Cisa
- Software
- Claude CodeorCodexorGitHub Copilot
Python
- Optional
- Web access
GitHub Copilot must be in agent mode.
Permissions
Allowed Hosts