Security audit
Security checks for vulnerabilities and agentic risk
This is a text-only paper-reading tutoring skill with disclosed local progress tracking that fits its purpose, though users should know it can save learning records in the workspace.
Install only if you are comfortable with the agent keeping local Markdown notes about your paper-reading performance, weaknesses, and session history under `.paper-reading-coach/`. Review or delete those files in shared, synced, or sensitive workspaces.
The skill explicitly instructs the agent to check for and persist learning-state/session data across conversations, but it does not require clear user notice or consent before reading or writing those local files. Because the stored content can include performance history, weaknesses, and prior session context, this creates a privacy risk through unexpected retention and reuse of user data.
The end-of-session workflow mandates saving updated learning-state and session-log files whenever writing tools are available, again without requiring informed user approval. In this skill’s context, the files can accumulate longitudinal profiles of the user’s reading ability and error patterns, making the privacy issue more significant because the retention is persistent, structured, and tied to future sessions.
The README states that the agent should automatically update workspace-level learning-state files if it has write access, but it does not clearly warn users that persistent local records about their performance and error patterns will be retained. This creates a privacy and consent issue: users may unknowingly leave behind sensitive study history, weaknesses, or paper-related notes on disk, especially in shared or synced workspaces.
This markdown file explicitly sets the session language to Chinese via "Language: 中文". Under the policy rule, a fixed language requirement without user choice or documented justification is a natural-language locale policy issue.
No suspicious patterns detected.
