Security audit
Security checks for vulnerabilities and agentic risk
The skill is a coherent error-recovery helper, but it can steer agents toward elevated permissions and alternate side-effecting upload paths without clear user approval or scoping.
Install only if you want a Chinese-first error-recovery workflow, and treat its remediation steps as suggestions that still need user approval before sudo, chmod, API uploads, credential changes, or other state-changing alternatives are used.
The trigger phrases are extremely broad and overlap with ordinary conversation about errors, failures, retries, and recovery. That makes unintended activation likely, causing the skill to inject its workflow into unrelated contexts and potentially steer the agent toward shell diagnostics or recovery actions when they were not requested.
The guidance explicitly suggests using sudo or modifying permissions as a generic recovery step. This normalizes privilege escalation during error handling, which is dangerous because many permission failures are intentional safety controls; bypassing them can expose sensitive data, damage system integrity, or train the agent to override least-privilege boundaries.
Referenced artifact was not completely inspected
node scripts/error-diagnose.mjs --error "ENOENT" --context "读取配置文件"
The file’s title, tagline, section headings, and instructions are presented in Chinese, with no indication that other languages are supported or that the locale restriction is intentional for a region-specific use case. This creates a natural-language policy concern because it imposes a specific language on users without opt-in or documented justification.
The skill documents shell-based actions and diagnostic commands but does not declare any tool scope or allowed-tools boundary. In practice this increases the chance that an agent invokes shell capabilities more broadly than intended during error handling, especially because recovery flows often run under time pressure and may touch sensitive files, network endpoints, or system state.
The description is written entirely in Chinese and presents the skill behavior in that language without offering any user language choice or documenting a justified region-specific limitation. This can violate language/locale policy when the skill is used in broader contexts where users have not opted into Chinese output.
The skill is framed as structured error recovery, but its recovery tree authorizes materially different operational actions such as switching from git push to direct GitHub API uploads and changing execution methods. That scope expansion can bypass normal review paths, alter data flow destinations, or cause the agent to perform side-effecting actions that the user did not explicitly authorize under the guise of recovery.
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.
quickFix: '检查chmod权限，或尝试sudo',
suggestions: [
'检查文件权限: ls -la',
'修改权限: chmod 644/600',
'切换到正确用户组',
'使用sudo（慎用）'
]
The single-word trigger '错误' is too generic and can activate the skill in many benign conversations. In this skill, accidental activation matters because the content includes operational recovery suggestions that may prompt unnecessary diagnostics or state-changing remediation behavior.
The trigger '失败' is common everyday language and can cause over-activation. Because the skill prescribes recovery steps and alternative execution paths, false activations can inappropriately alter the agent's behavior even when no real tool failure occurred.
The trigger '重试' is broad and may match ordinary discussion about trying again, not necessarily technical recovery. This can cause the skill to activate unexpectedly and push the agent into a retry framework that may not fit the task or safety context.
The trigger '恢复' is highly generic and likely to appear in many unrelated contexts. Given that the skill includes broad remediation ideas, unintended activation can expand the agent's autonomy in ways the user did not request.
The file's natural-language strings and usage documentation present the tool as a Chinese-language assistant, and all user-facing output is in Chinese. There is no indication that users can opt into another language or that the locale restriction is required for a region-specific purpose.
No suspicious patterns detected.