Security audit
Security checks for vulnerabilities and agentic risk
This is a GitHub CLI helper with mostly expected behavior, but it asks for write/exec authority and documents remote state changes without clear confirmation boundaries.
Install only if you are comfortable letting an agent run `gh` against your GitHub account. Use the narrowest GitHub scopes possible, avoid exposing private repository output unnecessarily, and require explicit approval before rerunning workflows, creating or changing GitHub objects, or setting persistent `gh` aliases.
The manifest and overview repeatedly position this skill as a free edition focused on '基础查询与检查能力' and '查询-检查-定位'. However, the documented commands include gh run rerun, which triggers a workflow rerun, and elsewhere gh alias set, which writes persistent local CLI configuration. These are state-changing actions beyond a purely query/check helper.
The documentation includes commands that can modify remote GitHub state, such as rerunning workflows, and says the capability supports creation operations, but it does not clearly warn users that these actions are state-changing. In an agentic exec context, lack of explicit confirmation boundaries can lead to unintended modifications to repositories, CI runs, or collaboration artifacts.
The trigger condition is overly broad, covering generic coding, debugging, testing, and deployment requests. In an agent environment with exec/write tools, this increases the chance the skill activates outside its narrow GitHub-management purpose and executes gh or shell commands in contexts the user did not intend, potentially causing unintended repository or workflow actions.
The '已知限制' section says the free version does not support '自动化工作流', but the examples provide gh alias set commands and a cross-repository batch query shell script. Those examples materially enable automation/persistent command customization, which conflicts with the claimed free-version limitation.
标题、摘要和全文说明均以中文固定呈现，且未说明是否支持根据用户偏好切换语言。按规则，若技能对语言/locale施加约束而没有用户选择或合理的区域性说明，可构成自然语言策略问题。
The skill instructs users to authenticate gh, use scopes including repo/workflow/read:org, and access private repositories, but it lacks explicit guidance on credential minimization, secret handling, and privacy boundaries. In an agent setting, this can normalize broad token use and increase the risk of unnecessary exposure of private repository data or over-privileged credentials.
The '已知限制' section explicitly states the free version does not support batch Issue/PR operations, yet the example directly below demonstrates a batch multi-repository Issue query script. Even though it is read-only, the documentation's blanket statement about batch Issue/PR operations is contradicted by the provided example.
No suspicious patterns detected.
