Security audit
Security checks for vulnerabilities and agentic risk
This is a domain research skill with some broad and external-reference behavior, but no hidden execution, persistence, or credential handling was found.
Before installing, users should understand that this skill may use named knowledge-base search heavily and may use external CBETA or document tools when the task calls for them. Avoid using it for private document editing unless you intend the agent to handle those files or online docs.
The trigger list is broad and includes common Tiantai-related terms that may appear in ordinary discussion, making unintended invocation more likely. Misfires can route user content into a more privileged or differently scoped workflow than expected, especially one that performs broad retrieval or uses external references.
The catch-all wording for 'any issue involving Tiantai doctrine' creates an unbounded scope that can absorb loosely related queries. Overbroad scope increases the chance of accidental invocation, unnecessary retrieval, and use of tools or behaviors the user did not intend.
The skill claims a constrained data-flow model of 'knowledge-base only' by default, but later introduces external CBETA APIs, online links, and online document tooling. This inconsistency can mislead operators and users about where data may be sent, weakening trust boundaries and increasing the chance of unintended external transmission or tool use.
The file requires use of Traditional Chinese for retrieval, including “必须用繁体中文检索,” without indicating user choice or documenting why this locale constraint should override user preference. This is a natural-language locale policy issue because it forces a language variant rather than offering opt-in or transparent justification in user-facing terms.
Adding local doc/docx processing and online document editing to a doctrinal research skill expands capabilities beyond the stated purpose, creating unnecessary attack surface. Extra tooling can enable data exfiltration, unintended file handling, or unauthorized modification workflows unrelated to answering religious research queries.
The skill includes a direct external API endpoint, which creates a clear path for transmitting prompts, search terms, or derived content outside the platform's internal knowledge bases. In a research skill that otherwise emphasizes KB-local retrieval, this external transmission path is more dangerous because users may not expect their queries to leave the trusted environment.
- Dev: https://cbdata.dila.edu.tw/dev/search?q=關鍵詞
- Stable: https://cbdata.dila.edu.tw/stable/search?q=關鍵詞
- 备用: https://api.cbetaonline.cn/search?q=關鍵詞
- 必须用繁体中文检索