Security audit
Security checks for vulnerabilities and agentic risk
This is a Markdown-only code analysis skill whose broad code reading, local reports, and optional command use fit its stated purpose, though users should be careful with archiving and any external LLM configuration.
Install only if you are comfortable letting the agent read the project directories you ask it to analyze and create .code-analyze reports/history. Keep archives out of shared repos if they contain sensitive code or findings, and do not configure ANALYZE_LLM_API_KEY or ANALYZE_LLM_ENDPOINT for proprietary code unless your organization approves that external service.
The documentation states that data stays on the local machine, but nearby sections show configuration for sending analysis to an external API endpoint. For a code-analysis skill handling potentially sensitive proprietary code, this is a serious transparency and privacy failure that can lead to unintended source-code exfiltration.
The skill exposes the exec tool but the description does not clearly warn users that it can execute shell commands. In an agent environment, hidden command-execution capability raises the risk of users invoking the skill without understanding it may modify files, read broad directories, or trigger external processes.
Automatic history archiving is enabled and described as a default behavior, but the documentation lacks a clear retention/privacy warning. For a code-analysis tool, archives may contain proprietary code snippets, findings, or security issues, creating unnecessary long-lived sensitive data on disk.
The skill claims local/offline historical storage while also documenting optional use of an external LLM API for batch analysis. This creates a misleading data-boundary expectation: users may assume analyzed code and history never leave the machine when batch workflows can transmit content to a third-party endpoint.
The documented external endpoint indicates the skill can be configured to send data to a remote API. In a code-analysis context, this may transmit sensitive source code, security findings, or internal architecture details to third parties if users enable the feature without understanding the privacy implications.
# 可选: 批量分析使用的外部 LLM Key(如非 Agent 内置)
export ANALYZE_LLM_API_KEY="your-api-key"
export ANALYZE_LLM_ENDPOINT="https://api.example.com/v1"
The display name and summary are presented only in Chinese, and the rest of the skill instructions are likewise written in Chinese without indicating that users may choose another language. This can violate a language/locale policy when no opt-in or locale justification is provided.