Security audit
Security checks for vulnerabilities and agentic risk
This skill is a file organizer whose core behavior is disclosed, but it asks agents to read, screenshot, rename, and move personal files with broad triggers, unclear rollback protection, and ambiguous privacy boundaries.
Review before installing. Use this only on explicitly chosen folders, start with dry-run, avoid important or sensitive directories unless you have backups, and confirm what model or API will see file previews, screenshots, OCR text, filenames, and metadata. Do not let it run automatically for generic file conversion or content extraction requests.
The trigger condition is overly broad, covering generic file handling, conversion, and content extraction requests beyond the skill’s stated purpose of visual file sorting. This increases the chance that an agent invokes a file-mutating skill in inappropriate contexts, potentially causing unintended renames, moves, or privacy exposure on unrelated tasks.
The description explicitly states '支持中文交互', and the document consistently presents the skill as Chinese-interaction-oriented without indicating that other languages are optional or that users can choose their preferred locale. Under the policy, forcing a specific language without opt-in is a natural-language policy violation unless clearly justified as region-specific.
The functional description highlights '中文交互' as a core capability, reinforcing a fixed language expectation rather than presenting language as a configurable user choice. This can conflict with organizational language/locale policy where users should be allowed to choose their interaction language.
The skill does not prominently warn up front that it will rename and move user files and that free-version changes may not be reversible. In a file-management context, lack of clear mutation/rollback warnings can lead users or agents to run it on important directories without understanding the risk of disruption or loss of traceability.
The skill’s safety claims are internally inconsistent: it advertises a hard prohibition on deletion, yet elsewhere states it supports delete operations. In an agentic file-management skill with exec/write capabilities, contradictory documentation can mislead users or downstream agents into authorizing destructive actions under a false sense of safety.
The skill promises local/private visual processing, but the dependency section requires an external LLM API, creating a privacy-boundary mismatch. Users may expose file contents, screenshots, or metadata to remote services while believing processing is local-only, which is especially risky for personal documents, invoices, and work files.
The manifest and main description frame the skill as a local visual file organizer, but the error handling instructs users to run ping and check firewall/proxy settings for network connectivity. Network-oriented behavior is not an obvious requirement for desktop file sorting as described, especially alongside claims of local/default visual processing.
