Security audit
Security checks for vulnerabilities and agentic risk
This looks like a simple skill skeleton generator, but its metadata and agent instructions materially misstate its behavior and required file access.
Install only if you are comfortable with a local generator that writes SKILL.md and AGENTS.md files. Run it in a new or empty output directory, because it can overwrite existing files. Treat the current metadata and AGENTS.md as unreliable until the publisher corrects the filesystem permission declaration and removes the analyzer-style PASS/FAIL contract.
The file presents contradictory behavior: it claims to generate a skill skeleton, but its output contract and rules describe a security/static-analysis tool that emits PASS/FAIL verdicts. This kind of specification mismatch can mislead an agent or operator into invoking the wrong capability, causing unsafe trust decisions, incorrect workflow automation, or bypass of expected review gates.
The manifest declares filesystem: false, but the skill description and usage clearly state that it generates SKILL.md, AGENTS.md, and directories on disk. This mismatch can mislead policy engines, reviewers, or users into granting trust under false assumptions, enabling unexpected file writes when the skill is invoked.
The manifest-like read_when activation guidance is written only in Chinese, which imposes a language-specific invocation/context for users despite the file otherwise being bilingual. This can violate language/locale policy expectations because no opt-in or alternative English trigger guidance is provided for activation conditions.
The top-level natural-language description presents the skill as '一键生成' with Chinese-first wording and does not indicate that users can choose their preferred language or locale. Under the policy rule, language-forcing text without opt-in can be a locale-policy concern even when the code itself is functional.
No suspicious patterns detected.
