Security audit
Security checks for vulnerabilities and agentic risk
This skill is a straightforward chat-to-Markdown exporter that saves user-reviewed conversation summaries locally.
Install only if you are comfortable with selected conversation content being written to local Markdown files. Review the preview and destination path carefully, use dry-run or cancel for sensitive chats, and avoid exporting secrets or private details unless the storage location is appropriate.
The skill exports conversation content to disk but does not warn users that potentially sensitive chat data may be persisted locally. This creates a real privacy risk because users may unknowingly save secrets, personal data, or confidential project details into files that could later be accessed, synced, or backed up elsewhere.
The skill explicitly scans prior conversation data for previous conversation_id values and supports updating or appending to earlier exports, which introduces session persistence and cross-session linkage. This can make separate conversations traceable over time and increase privacy exposure by correlating discussions that a user may have expected to remain isolated.
Before generating a new file, scan the conversation for a previous `conversation_id` for the same topic.
- If no previous export exists, generate a new ID
- If a previous export exists, ask whether to update, append, or create a new KB
Conversation ID format:
Hardcoding language: en without user opt-in can cause the exported knowledge base to misrepresent the conversation language or coerce multilingual/private content into English-oriented output. While not a direct code-execution issue, it can lead to inaccurate records, privacy misunderstandings, and unintended transformation of user content.
The description is broad enough to match many ordinary user requests such as saving, exporting, or documenting a conversation. In agent ecosystems that auto-route by natural-language matching, this can cause the skill to trigger unexpectedly and gain access to conversation content outside the user's precise intent, increasing the risk of unintended data capture or exfiltration into files.
No suspicious patterns detected.