CVSS v4.0 Scoring turns the text of a vulnerability finding into a vector string. It extracts the security-relevant facts first, then scores the exploitability metrics (attack vector, complexity, requirements, privileges, user interaction) and the impact metrics for the vulnerable and subsequent systems, using a metrics reference file for definitions.
It does not compute the numeric score. It returns the vector plus a FIRST.org calculator link with the vector applied, flags uncertain impact values as ASSUMED, and refuses input that does not describe an exploitable weakness.
What You Get
- A CVSS v4.0 vector string for each finding
- A FIRST.org calculator link with the vector pre-filled, where the numeric score comes from
- ASSUMED flags on impact metrics that the description left open
- A visible note when the input used CVSS v3.1 wording and was translated
Common Use Cases
- Scoring findings for a security audit report
- Estimating relative severity across several bugs
- Migrating v3.1 scores to v4.0 and catching the Scope and User Interaction differences
- Answering a developer asking what the CVSS is for a bug
What Makes It Unique
It targets v4.0 specifically and lists the reasoning errors that skew scores, such as treating network exposure as attack vector, marking every impact high, counting a dependency as a subsequent system, or putting prerequisites in complexity. It also sends the arithmetic to the official calculator and does not attempt it.
Under The Hood
The skill is one instruction file plus metrics and example references, and its only declared tool is Read. The instructions run no commands, write no files and make no network requests; the calculator link is text for you to open. The agent reasons over your description using the metric definitions and returns the vector in chat. The folder also holds a hidden recipe file for the separate calcifier runtime. It defines a structured assessment format that a worker agent fills in and that is rendered deterministically.