Security audit
Security checks for vulnerabilities and agentic risk
This is mostly a Git help skill, but its activation scope is much broader than Git and it includes command-capable, persistent, and destructive Git operations without tight user-control guidance.
Install only if you want a Chinese-language Git command helper and are comfortable with an agent suggesting or running Git shell commands. Review commands before execution, especially git clean, git reset, git branch -D, git stash clear, git push, and git config --global; avoid letting the skill activate for non-Git coding, planning, deployment, or team-management requests.
The trigger condition includes broad project management, task planning, progress tracking, and team collaboration scenarios, which materially exceed the skill's stated Git-focused purpose. Over-broad activation criteria increase the chance that the agent invokes a command-capable skill in inappropriate contexts, expanding attack surface and enabling unintended repository or shell operations.
The manifest and description position the skill as a Git workflow helper, but the invocation guidance expands it into generic code generation, debugging, testing, and deployment. This scope inflation can cause the agent to activate the skill in unrelated development contexts and grant command-execution behavior beyond what users would reasonably expect from a Git-only assistant.
The description tells users to use the skill for broad developer activities like code generation, programming assistance, debugging, testing, and deployment. This overlaps with many general requests and can cause the agent to over-select a shell-capable skill for tasks that do not require Git operations, increasing the risk of unintended command execution.
The manifest description states '支持中文交互', and the document consistently presents the skill as Chinese-only without an explicit opt-in or alternative language option. This can violate language/locale choice expectations because the skill appears to impose a locale preference rather than offering it to the user.
The examples include destructive cleanup commands like 'git clean -f' and 'git clean -fd' without strong, immediate warnings about irreversible deletion of untracked files. In an exec-enabled skill, users or agents may copy these commands directly, causing unintended data loss in local repositories or working directories.
The document claims the tool is 'pure Markdown instruction driven' and needs no extra API key, yet elsewhere it explicitly supports exec, remote push, HTTPS, and network troubleshooting. These inconsistent trust signals can mislead operators about the skill's execution and network capabilities, reducing scrutiny around command execution and remote interactions.
No suspicious patterns detected.
