Security audit
Security checks for vulnerabilities and agentic risk
This meeting knowledge-base skill appears purpose-built and not malicious, but it can aggregate sensitive meeting history into a persistent searchable file with broad natural-language triggers and limited scoping guidance.
Review before installing if your Tencent Meeting history or local notes contain confidential business information. Use it only on explicitly chosen meetings or directories, keep the generated knowledge-base file in an access-controlled location, and delete or redact outputs that should not remain searchable.
The trigger phrases are broad conversational requests such as building a knowledge base, checking prior decisions, or summarizing recent meeting conclusions. In an assistant environment, these overlap with normal user intents and can cause unintended activation of the skill, which may pull or process meeting data when the user did not explicitly mean to invoke this specific capability.
The skill describes local script behavior that reads meeting materials and writes a knowledge-base output, but it does not declare any explicit tool scope or permission boundaries. In an agent setting, this ambiguity can lead to overbroad file access, unintended reading of sensitive local documents, or writing searchable aggregates without clear user consent or sandbox restrictions.
The skill centers on aggregating multiple meeting transcripts and notes into a searchable knowledge base, but it does not present a prominent user-facing warning about the sensitivity, retention, and discoverability implications of that aggregation. Cross-meeting indexing materially increases privacy and confidentiality risk because dispersed internal discussions, decisions, owners, and risks become easy to search and reuse across contexts.
Several example phrases such as building a meeting knowledge base, checking prior decisions, or summarizing meeting conclusions are generic requests that overlap with normal assistant usage. Because the content domain is highly sensitive—internal meetings, decisions, owners, risks, and action items—mis-triggering could expose or consolidate confidential organizational information beyond the user's intended scope.
The skill says users can speak in everyday language and that the AI will automatically recognize and invoke the skill, which creates a broad trigger surface. This can cause unintended activation during ordinary conversation, leading to accidental retrieval, aggregation, or persistence of sensitive meeting content when the user did not explicitly intend to use this skill.
The trigger phrases are broad natural-language requests such as building a knowledge base, checking what was decided earlier, or summarizing recent meetings, which can overlap with ordinary conversation in a meeting or assistant context. This can cause unintended activation of the skill on unrelated user input, potentially pulling together sensitive multi-meeting notes and exposing or processing data the user did not explicitly intend to access through this skill.
All user-facing instructions and examples in the file are Chinese-only, and the skill does not indicate that language selection is optional or limited to a justified region-specific context. This may violate language/locale policy if users are not given an explicit choice.
This Python file contains user-facing docstrings, CLI help text, status messages, and output content entirely in Chinese, which imposes a specific language/locale on users. The policy allows locale constraints only when documented and justified or when users are given a choice; this file provides neither.
