菜单
Evaluate raw non-functional test results against the NFR catalog. Use when user asks to "score NFR results", "compare SLOs", or "decide if the release passes non-functional gates". Outputs pass, degraded, or fail verdicts with evidence links for every requirement.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/cyberskill-official/cyberos --skill nfr-evaluator命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Audit a hardening-record@1 for scope discipline, verification honesty, and gate integrity. Checks every changed file was inside a worked finding's declared scope, every closed finding carries verbatim validation output and a regression gate, fingerprints were carried unchanged, and no finding closed on operator silence. Use when asked to "audit this hardening session", "check what /harden actually did", or "verify the remediation record". Do NOT use to work findings (harden-record-author) or to ship class:improvement tasks (ship-tasks).
Work findings from an inspection-report@1 under scope and safety discipline and emit a hardening-record@1. Validates input with inspect-lint before reading findings, classifies each finding as agent/operator/split (honouring operator_prerequisites), halts at plan and review gates, and never pushes, merges, rotates credentials, or executes irreversible actions. Use for "/harden", "work the inspection backlog", or "remediate what /inspect found". Do NOT use to discover findings (inspection-report-author), judge report quality (inspection-report-audit), or ship a backlog class:improvement task (ship-tasks — distinct "harden a task" path).
Run the CyberOS ship-tasks workflow in any repo: drive a task (product or improvement class) through implement -> review -> test -> done, with HITL required at the two human-acceptance gates and gates from the repo's own build/lint/test. Use when asked to ship, implement, or harden a task (backlog class:improvement), or to drive a docs/tasks backlog. Do NOT use for full-repo inspection remediation — that is /inspect then /harden on an inspection-report@1.
基于 SOC 职业分类
正在显示 SKILL.md
| name | nfr-evaluator |
| description | Evaluate raw non-functional test results against the NFR catalog. Use when user asks to "score NFR results", "compare SLOs", or "decide if the release passes non-functional gates". Outputs pass, degraded, or fail verdicts with evidence links for every requirement. |
| metadata | {"version":"1.0.0","module":"skill"} |
| allowed_memory_scopes | {"read":["project:*"]} |
| allowed_mcp_tools | ["audit.append"] |
Compare test output to SLO thresholds and produce one verdict per NFR.