Async Programming

ReviewAudited by ClawScan on May 10, 2026.

Overview

This instruction-only skill is review-worthy because it automatically launches run-mode coding subagents for broad programming tasks without clear approval, cancellation, sandboxing, or rollback rules.

Install only if you want coding requests to be delegated automatically to background subagents. Use it on version-controlled or sandboxed projects, avoid sharing secrets in task descriptions, monitor spawned tasks, and review diffs before relying on any generated changes.

Findings (3)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

A normal coding request could start an autonomous task that changes project files or setup before the user has reviewed an exact plan.

Why it was flagged

The skill directs the agent to launch a run-mode subagent for broad programming tasks that may modify code, configuration, or dependencies.

Skill content
当用户请求涉及以下内容时,激活此技能:... 编写/修改代码 ... 项目配置 ... 依赖安装 ... sessions_spawn({ ... mode: "run", runTimeoutSeconds: 900 })
Recommendation

Require explicit confirmation for write/install actions, scope each task to a specific project path, and ask the subagent to present a diff or plan before applying changes.

What this means

Multiple agents could keep working on projects in the background, making overlapping or unexpected changes while the user continues chatting.

Why it was flagged

The core workflow is to keep background subagents running asynchronously, including parallel execution, while the main chat continues.

Skill content
调用子 agent 后,立即回复用户,不要等待完成! ... 最多 8 个子 agent 同时运行
Recommendation

Add visible task controls such as approve, pause, cancel, status, and per-task boundaries; avoid parallel writes to the same project unless the user explicitly opts in.

What this means

Project names, paths, and task details may be shared with the child agent/model during delegation.

Why it was flagged

The spawned subagent/model receives task details including project paths and requirements, which is expected for the purpose but is still an inter-agent data flow.

Skill content
task: "{具体任务描述,包含项目路径和具体要求}", ... model: "bailian/qwen3-coder-plus"
Recommendation

Avoid including secrets or sensitive code details in delegated task descriptions unless you are comfortable sharing them with the configured subagent/model provider.