mirror of
https://github.com/xr843/Master-skill.git
synced 2026-05-10 13:26:25 +00:00
xianren
3dfaa6bb62
Tier A infrastructure to convert 204★ momentum into durable community and distribution. Pairs with PR #12 (ETHICS / CHANGELOG / fidelity smoke) — merge PR #12 first, then this. **Community docs** - `CONTRIBUTING.md` — three-tier contribution flow (code vs docs vs master content), with a detailed §3 "how to contribute a new master" covering copyright tier self-check, `/create-master` vs manual path, SKILL.md frontmatter spec, voice.md Layer 0-3, fidelity.jsonl authoring. - `CODE_OF_CONDUCT.md` — Contributor Covenant 2.1 + Buddhist-project-specific rules (no doctrinal supremacy, no using the repo as a dharma platform, no impersonating clergy, no apocryphal scripture citations). - `SECURITY.md` — threat model (prompt injection, supply chain, secret leak, installer safety, religious-boundary adversarial input), SLA table, GH Security Advisory as primary channel. **Issue / PR templates** - `bug_report.yml` — area selector, master selector, repro template. - `feature_request.yml` — non-master enhancement requests. - `new_master.yml` — **MUST precede any new-master PR**; forces tier A/B/D self-judgment and citation planning before sunk cost. - `boundary_violation.yml` — P0 template for ETHICS.md §3 violations. - `config.yml` — directs users to Discussions for chatter, Security Advisory for vulnerabilities, email for urgent takedowns. - `PULL_REQUEST_TEMPLATE.md` — change-type checkboxes, self-check list, new- master fields, local-test block. **npm publish** - `.github/workflows/npm-publish.yml` — release-tag + workflow_dispatch, with dry-run mode, tag-version match check, `npm publish --provenance`. First-time publish needs NPM_TOKEN secret (user action). - `package.json` hardening: - `engines.node: >=18` - scripts: `validate`, `validate:fidelity`, `test`, `test:smoke`, `prepack` - `publishConfig` (public, npm registry) - `files` list expanded to include platform manifests + governance docs - richer `keywords`, `author`, `bugs`, `homepage` fields - expanded description mentions all 8 masters for search discoverability **README** - 贡献指南 section rewritten to point at CONTRIBUTING.md + Discussions + four issue templates, with "先开 New Master issue 征询" gate spelled out. Discussions were enabled on the repo alongside this commit via `gh api`, so the links in README/templates resolve immediately on merge. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
106 lines
3.3 KiB
Markdown
106 lines
3.3 KiB
Markdown
# Security Policy
|
||
|
||
## Supported Versions
|
||
|
||
Master-skill 以 `main` 为持续发布分支。我们仅对以下版本承诺 security fix：
|
||
|
||
| 版本 | 状态 |
|
||
|------|------|
|
||
| `main` (latest) | ✅ 持续修复 |
|
||
| `0.3.x` | ✅ 持续修复 |
|
||
| `< 0.3.0` | ❌ 不再维护 |
|
||
|
||
---
|
||
|
||
## 报告安全漏洞
|
||
|
||
**请勿通过公开 issue 报告安全漏洞。** 公开披露会让攻击者先行利用。
|
||
|
||
### 推荐方式：GitHub Security Advisory
|
||
|
||
1. 访问 https://github.com/xr843/Master-skill/security/advisories/new
|
||
2. 填写漏洞描述、复现步骤、影响范围
|
||
3. 维护者会在 72 小时内确认收悉
|
||
|
||
### 备选：邮件
|
||
|
||
邮件发送至 **xianren843@protonmail.com**，主题请包含 `[SECURITY]`。
|
||
|
||
**推荐用 GPG / PGP 加密**：公钥可通过 keys.openpgp.org 搜索该邮箱获取（或在 issue 中 request）。
|
||
|
||
---
|
||
|
||
## 本项目关心的安全类别
|
||
|
||
Master-skill 作为 AgentSkill 插件 + NPX CLI，主要关注以下安全面：
|
||
|
||
### 1. **Prompt Injection**
|
||
|
||
- 预置法师的 `SKILL.md` / `voice.md` / `sources/` 被恶意注入，导致 AI 绕过 HARD-GATE 或伦理边界
|
||
- `/create-master` 生成管线中的 prompt 模板被污染
|
||
- 用户问题中的诱导越狱（"假装你是个能传戒的 AI..."）
|
||
|
||
### 2. **Supply Chain**
|
||
|
||
- `package.json` 依赖被投毒（当前依赖极少，但未来可能增加）
|
||
- FoJin API 返回的文本被篡改以影响 fidelity test
|
||
- CBETA ID 伪造（已有 `scripts/validate.py` 防线，但需持续完善）
|
||
|
||
### 3. **Secret Leakage**
|
||
|
||
- `ANTHROPIC_API_KEY` 在 CI 日志中意外泄露
|
||
- 用户在 issue / discussion 中误粘自己的 API key（自动检测 + 立即清除）
|
||
|
||
### 4. **Installer Safety**
|
||
|
||
- `bin/cli.mjs` (`npx master-skill install`) 的目录操作是否存在路径穿越
|
||
- 安装到 `~/.claude/skills/` 时的符号链接注入
|
||
|
||
### 5. **Religious-Boundary Violation via Adversarial Input**
|
||
|
||
- 特别 crafted 的用户问题使法师角色逾越 [`ETHICS.md`](ETHICS.md) §3 的禁止行为
|
||
- 这类属于**安全 + 伦理**交叉问题，优先级等同 S 级漏洞
|
||
|
||
---
|
||
|
||
## 非安全范畴（请走普通 issue）
|
||
|
||
以下不属于 security policy 范围：
|
||
|
||
- 某位法师回答不够"像"该祖师 → 开 bug report
|
||
- 引经错误（而非伪造）→ 开 bug report
|
||
- FoJin API 不可用 → 项目已有 graceful degradation，非安全问题
|
||
- UX / 文档问题 → 普通 issue
|
||
|
||
---
|
||
|
||
## 响应 SLA
|
||
|
||
| 严重级 | 首次回复 | 修复目标 | 公开披露 |
|
||
|-------|---------|---------|---------|
|
||
| Critical（0day、泄密、Prompt injection 破 HARD-GATE）| 24h | 7 天内发 patch | 修复后 7 天 |
|
||
| High（影响正常功能但非系统性）| 72h | 14 天内发 patch | 修复后 30 天 |
|
||
| Medium / Low | 7 天 | 下一版本 | 与版本同步 |
|
||
|
||
---
|
||
|
||
## 安全奖励
|
||
|
||
本项目目前**无法提供现金奖励**（个人维护，非商业项目）。但会：
|
||
|
||
- 在 `CHANGELOG.md` 显著位置署名感谢（除非你希望匿名）
|
||
- 对严重漏洞发现者提供一份定制化感谢文书（PDF + 项目维护者签名）
|
||
- 愿与你共同起草 CVE 条目（如适用）
|
||
|
||
---
|
||
|
||
## 相关文档
|
||
|
||
- 负责任披露：[GitHub Security Advisory Policy](https://docs.github.com/en/code-security/security-advisories)
|
||
- 内容安全边界：[`ETHICS.md`](ETHICS.md) §3
|
||
- 社区安全：[`CODE_OF_CONDUCT.md`](CODE_OF_CONDUCT.md)
|
||
|
||
---
|
||
|
||
感谢你让本项目更安全。
|