AGENT SKILL

security

Security pattern database for Claude Code plugin audits. Covers execution surfaces, supply chain, data exfiltration, and prompt injection.

License 许可
MIT
最近更新
2026-11-01
Stars
★ 3
Forks
1

01 前言 Front Matter

name
security
version
0.1.0
description
Security pattern database for Claude Code plugin audits. Covers execution surfaces, supply chain, data exfiltration, and prompt injection.

03 SKILL.md 正文

Security Scan Patterns for Claude Code Plugins

Execution Surfaces

Claude Code plugins have five execution surfaces that must be scanned:

Surface Files Risk Level Why
Hooks hooks/hooks.json, referenced scripts Critical Runs on EVERY tool call automatically
Scripts scripts/*.sh, *.py, *.js High Executed by commands/agents
MCP Servers .mcp.json High Network access, data flow
Bash in commands commands/*.md with Bash tool Medium Shell execution via Claude
Install scripts package.json postinstall, setup scripts Medium Runs on install

Dangerous Shell Patterns

Critical (immediate risk)

Pattern Regex Why
Pipe to shell curl.*|.*sh, wget.*|.*bash Remote code execution
Eval with variables eval\s+["']?\$ Arbitrary code execution
Reverse shell bash\s+-i\s+>&, /dev/tcp/ Backdoor
Base64 decode and exec base64.*|.*sh, base64.*|.*python Obfuscated execution
SSH key exfiltration cat.*\.ssh/, scp.*\.ssh/ Key theft
Token exfiltration Secrets like GITHUB_TOKEN or API keys sent to curl/wget Credential theft

High (likely dangerous)

Pattern Regex Why
Subprocess with shell=True subprocess\.(call|run|Popen).*shell\s*=\s*True Python shell injection
OS system calls os\.system\( Python shell execution
Dynamic require/import require\(\s*\$, import\(\s*\$ Dynamic code loading
File write outside repo > ~/, > /etc/, > /tmp/.*\.sh System modification
Sudo usage sudo\s+ Privilege escalation
PATH modification Appending to bashrc, zshrc, or profile Persistent system modification

Medium (context-dependent)

Pattern Regex Why
Network calls curl\s+, wget\s+, fetch\(, requests\.(get|post) Data exfiltration potential
Environment access process\.env, os\.environ, shell variable expansion Sensitive data access
File reads outside repo Reading from home directory or system paths Data access
Runtime package install npm install, pip install, gem install Supply chain risk
Shell exec functions Functions that execute strings as shell commands Shell execution

MCP Configuration Risks

Scan .mcp.json for:

Risk Check Severity
Remote servers url field pointing to non-localhost High
Unknown domains Domain not in known-safe list High
Broad permissions permissions with wildcard or extensive list Medium
File system access Server with fs or filesystem capability Medium
Shell access Server with shell or execution capability Critical
Missing auth Remote server without auth field High

Known-safe MCP domains: localhost, 127.0.0.1, modelcontextprotocol.io, github.com, api.anthropic.com

Hook Safety Rules

Scan hooks/hooks.json for:

Risk Check Severity
Hook runs shell script command field references .sh, .py, .js Medium (must scan the script)
Hook uses user input Script receives prompt or input variables without sanitization High
Hook on every event Triggers on PreToolUse or PostToolUse without tool filter Medium
Hook modifies files Script writes to disk on every tool call Medium
Hook makes network calls Script contains network request commands High

Dependency Supply Chain

Scan package.json for:

Risk Check Severity
postinstall scripts scripts.postinstall exists High
preinstall scripts scripts.preinstall exists High
Git URL dependencies Deps pointing to git URLs Medium
Unpinned versions Wildcard or "latest" version Medium

Scan requirements.txt / pyproject.toml for:

Risk Check Severity
Git URL deps git+https or git+ssh URLs Medium
Unpinned No version pin Low
Direct URL HTTP download URLs High

Prompt Injection Surfaces

Risk Check Severity
Untrusted file content in prompts Agent reads arbitrary file then uses content in Bash High
User input passed to shell Command takes arguments and passes to Bash without sanitization Critical
Template expansion Variable expansion in hook scripts with user-controlled values High

Severity Definitions

Severity Meaning Action
Critical Immediate exploitation risk: RCE, credential theft, backdoor Block contribution, file security issue
High Likely dangerous: shell injection, data exfil, privilege escalation Block contribution, report in audit
Medium Context-dependent: network calls, env access, runtime installs Report in audit, flag for review
Low Minor concern: unpinned deps, broad permissions Report as informational

Report Format

The security scan section in an audit report follows this structure:

## Security Scan

| Severity | Count |
|----------|-------|
| Critical | N |
| High | N |
| Medium | N |
| Low | N |

### Findings

| # | Severity | File | Line | Pattern | Description |
|---|----------|------|------|---------|-------------|

Risk Gate

If any Critical or High findings exist, the contribute-approved label must NOT be applied. The audit report must include a prominent warning and the tracking issue must link to the security findings.

04 同仓库更多 Skills

shengdabai
shengdabai/Tony-Claude-Code-Skills

seedance-prompt-en

Write effective prompts for Jimeng Seedance 2.0 multimodal AI video generation. Use when users want to create video prompts using text, images, videos, and audio inputs with the @ reference system. Covers camera movements, effects replication, video extension, editing, music beat-matching, e-commerce ads, short dramas, and educational content.

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

onepage-pdf

Convert an HTML page into a single continuous-page PDF (one tall page, no pagination breaks), preserving desktop layout, backgrounds and selectable text. Use when the user asks to turn an HTML file/report/proposal into a "single-page PDF", "long PDF", "不分页 PDF", "单页 PDF", "长图式 PDF", or complains that an HTML-to-PDF export breaks into pages or cuts content. Supports optional string redaction with leak verification.

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

native-feel-cross-platform-desktop

Use when the user is designing, prototyping, or rewriting a desktop app that must run on multiple OSes (macOS + Windows, optionally Linux) AND feel indistinguishable from a native app to its users — fast launch, native windowing, native input handling, native materials. Trigger words include "cross-platform desktop", "Electron alternative", "Tauri vs native", "WebView wrapper", "near-native performance", "Raycast architecture", "WebKit/WebView2 quirks", "WKWebView", "system tray app", "global hotkey app", "launcher app". Do NOT trigger this skill for pure web apps, pure mobile apps, or for greenfield projects that have no native-feel requirement.

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

gstack

Router for the gstack skill suite. (gstack)

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

fireworks-tech-graph

Use when the user wants to create any technical diagram - architecture, data flow, flowchart, sequence, agent/memory, or concept map - and export as SVG+PNG. Trigger on: "画图" "帮我画" "生成图" "做个图" "架构图" "流程图" "可视化一下" "出图" "generate diagram" "draw diagram" "visualize" or any system/flow description the user wants illustrated.

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

dbs-video-extract

接收抖音、小红书、微信视频号等短视频链接或分享文案,查询作品/账号数据并提取语音文字稿,按作者和标题归档为 Markdown。用户希望完整解析短视频、取得数据或文字稿时使用;运行前检查 TikHub 与轻抖使用凭证,缺少时用新手能理解的方式引导购买和安全配置。

★ 3