ClawHub

Claude Code Patterns

v1.0.0

Claude Code 源码中提取的 AI Agent 工程模式。当需要设计 agent 调度、工具并发执行、上下文压缩、状态机循环、流式处理等复杂 agent 系统时使用此 skill。触发场景：(1) 设计多轮对话 agent (2) 实现工具并发执行 (3) 构建上下文管理策略 (4) 优化 agent 启...

⭐ 0· 57·0 current·0 all-time
bysirius@siyrs
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
✓
Purpose & Capability
The name/description (agent engineering patterns) aligns with the included SKILL.md and reference documents. All examples and patterns are consistent with designing agent loops, streaming executors, compaction, and error recovery; nothing in the metadata asks for unrelated privileges or credentials.
ℹ
Instruction Scope
The SKILL.md and reference files are purely documentation and code examples. They do reference concepts like 'keychain prefetch', 'MDM config', and executing Bash tools as examples — which are appropriate for patterns but could imply sensitive operations if a developer implements them. The documents do not instruct the platform agent to read system files or environment variables at install/runtime.
✓
Install Mechanism
No install spec and no code files that would be executed at install time. Instruction-only skills are lowest-risk from an install perspective.
✓
Credentials
The skill requires no environment variables, credentials, or config paths. Example snippets mention fetching keychain or MDM in the pattern context, but the skill itself does not request or require secrets.
✓
Persistence & Privilege
always is false and the skill is user-invocable only. There is no mechanism provided to persist or modify other skills or system-wide settings.
Assessment
This skill is documentation and example code for building agent systems — it appears internally consistent and does not ask for credentials or install components. If you plan to adopt these patterns in real software, be careful when implementing examples that touch keychains, MDM, or run Bash: those concrete implementations can access sensitive data or execute commands. Before using these patterns in production, review any concrete tool implementations for: (1) safe handling of secrets (don’t prefetch keychains into unprotected memory), (2) validation and sandboxing of Bash/OS commands to avoid arbitrary execution, (3) concurrency and cancellation behavior to prevent race conditions or unintended side effects, and (4) any external API calls for compaction/summary models to ensure tokens/data are handled appropriately.

Like a lobster shell, security has layers — review code before you run it.

latestvk97bh1z532mr4ccncm6vxjm7fn84599g

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Comments