Use when 要判一則 Codex Security finding 是真漏洞還是假警報(finding mode),或上線前要分清 repo / staging / production 各層還缺哪些安全證據(map mode)。NOT for 修 code、跑掃描、寫 SECURITY.md。
Scanned 9/5/2026
Install to Claude Code
npx -y skills add Charles5277/nuxt-supabase-starter --skill security-evidence --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Security Evidence?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/charles5277-security-evidence-nuxt-supabase-starter)More formats (shields.io, HTML) on the badges page.
---
name: security-evidence
description: Use when 要判一則 Codex Security finding 是真漏洞還是假警報(finding mode),或上線前要分清 repo / staging / production 各層還缺哪些安全證據(map mode)。NOT for 修 code、跑掃描、寫 SECURITY.md。
effort: high
permission_tier: read
---
<!--
🔒 LOCKED — managed by clade
Source: plugins/hub-core/skills/security-evidence/
Edit at: $CLADE_HOME
Local edits will be reverted by the next sync.
-->
# security-evidence
兩個 mode,各對應一份逐字 prompt(`references/`)。**MUST 先讀 `modes.md` 判 mode,再讀對應 reference 全文照它的 context-gathering 逐步問**——它們刻意分小步等答案,不要一次把全部問題丟出去。
| mode | 什麼時候 | reference | 產物 |
| --- | --- | --- | --- |
| `finding` | 一次一則 finding;報告出現 High / Medium、看不懂、有爭議、修之前 | `references/finding-evidence-explainer.md` | Finding Evidence Review:Source / Control / Sink、反證、Severity 與 Confidence 分開、Coverage、Proof Gaps、verdict(accept / needs more validation / unsupported) |
| `map` | 上線前、重大改版、修完重要 finding、掃描回 No findings | `references/production-blind-spot-mapper.md` | Production Security Evidence Map:五層 Coverage、P0 / P1 / P2 清單(pass condition、evidence、owner)、scope-limited verdict |
## 讀取順序(唯讀)
1. target 的 `SECURITY.md`(安全憲法)——每一條 `INV-n` 都是判 counterevidence 的依據;沒有這份就先說明 Confidence 會被壓低
2. finding 原文(`<output_dir>/findings.json` 該條 + `report.md` 對應段)與 `coverage.json` 的 Coverage
3. 被點名的檔案及其直接 caller / callee;**NEVER** 無方向爬 codebase
## 出口
- verdict `accept` / `needs more validation` → 登 TD(consumer 自家 `docs/tech-debt.md`),`### 自驗` 寫 Proof Gap 的驗證動作;修完跑 `scripts/security-scan.ts verify --finding <id>`
- verdict `unsupported` → 寫進本次 review 報告,不開 TD、不改 code
- `map` 的 P0 / P1 每一項各登一條 TD,owner 缺就寫 `Owner Missing`,**NEVER** 自己指派
## NEVER
- NEVER 把 scanner 的 High 標籤當成 Severity 與 Confidence 都是 High——兩者分開判
- NEVER 因為 `No findings` 就給 READY 或「安全」
- NEVER 要 secret 值或真實客戶資料;要 key 名與遮蔽片段
- NEVER 對 production 發測試流量或改設定;本 skill 只產出判讀與驗證計畫
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!