Security audit
Security checks for vulnerabilities and agentic risk
The skill is a disclosed local task-gamification add-on, but users should review its persistent tracking and integration edits before installing.
Install only if you want the agent to maintain persistent local gamification state for your tasks and to integrate that state into briefings or cron prompts. Before applying integration changes, confirm what deactivation means, review the AGENTS.md/HEARTBEAT.md/cron diffs, and treat the Korean templates as a locale choice you may want to adapt.
The skill content and prescribed user-facing messages are primarily written in Korean, with mixed English terms, and the document does not state that language should follow the user's preference. This can violate a language/locale policy when the skill implicitly forces a specific language for responses without opt-in.
The skill contains conflicting instructions for inactive mode: one section says tracking should continue while deactivated, while Best Practices says all quest-state updates must stop when active: false. This ambiguity can cause the agent to update user state contrary to expectations, leading to privacy/consent violations and inconsistent behavior across sessions.
The theme-change trigger mapping is broad enough that ordinary conversational words like '학교', '연구', or '우주' could be interpreted as commands to change the user's theme even when the user is only discussing task content. In a task-gamification skill that updates persistent state automatically, this can cause unintended state changes and user confusion, though it is not code execution or direct data exfiltration.
The document makes a misleading safety claim that the skill does not directly modify workspace files, but later instructs the agent to update task-quest/quest-state.md, task-quest/achievements.md, and manually edit AGENTS.md and HEARTBEAT.md. This contradiction can cause users or downstream agents to grant broader trust than warranted, leading to unauthorized or insufficiently reviewed file modifications in the workspace.
The guide states that task data is read-only and implies there are no writes, but it also directs automatic XP calculation and updates to quest state on task completion. Even if writes are limited to task-quest/, the inconsistent description obscures the skill's real side effects and may cause operators to enable automation without understanding that task-derived state will be persisted.
This shell script initializes user-facing markdown content with Korean-only headings and guidance such as the title and recent-activity text. It also prints a Korean trigger phrase at the end, but nowhere indicates that the skill is intentionally Korean-only or gives the user a language choice, which can violate language/locale policy requirements.
The achievements and history templates include multiple Korean-only section headers, table labels, and instructional text. Because these files are created automatically for the user, the script is effectively enforcing a locale in generated artifacts without stating that the skill is region-specific or configurable.
This markdown file presents all skill instructions and mechanics in Korean, and does not indicate that the user can choose another language or locale. The policy for natural-language violations applies to all file types and flags forced language constraints without explicit opt-in or justification.
The natural-language instructions, examples, and output messages are all written in Korean, which can imply a fixed-language behavior. There is no indication that users may choose another language or that the Korean-only presentation is an intentional, documented locale constraint.
