Security audit
Security checks for vulnerabilities and agentic risk
This is a small bulk text replacement skill, but its permission metadata understates the file and command access it actually needs.
Install only if you are comfortable with a local bulk-editing tool that can rewrite every file matched by its glob when --apply is used. Preview first, use narrow glob patterns, and treat the package metadata as needing correction because it says no filesystem/subprocess permission even though the skill needs local file and CLI execution access.
The manifest declares filesystem: false while the skill’s documented purpose, allowed tools, and workflow explicitly require reading and writing files and creating .bak backups. This mismatch can bypass policy expectations, cause incorrect sandboxing or approval decisions, and mislead an agent into performing file-modifying actions under a supposedly no-filesystem capability declaration.
The file’s instructional content for AI agents is entirely in Chinese, including operational guidance and constraints. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale constraint is explicitly justified.
The safety section says the tool 'moves / renames only' even though the rest of the skill describes in-file bulk text replacement. Misstating the operation type can cause operators or automated reviewers to underestimate the risk of content modification, leading to accidental destructive edits across many files.
The file presents the skill content in fixed Chinese and English sections and labels, but does not indicate that language is optional or user-selectable. Under the policy rule, language constraints should either offer opt-in/choice or be clearly justified as region-specific.
The top-level manifest description is written only in Chinese, while later sections are bilingual. This creates a language preference in user-facing metadata without indicating user choice or an opt-in mechanism, which can violate language/locale policy expectations for neutral skill presentation.
The skill's primary display_name field is Chinese-only, even though an English alternative exists separately. Using a single-language primary label can amount to a locale preference unless the system explicitly selects names based on user settings.
