T05 · Unauthorized Access and Privilege Escalation
- Location
SKILL.md:75- Finding
Overbroad Cross-Session History and Shared-Memory Access
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
The skill is transparent about syncing memory, but it asks for broad access to prior sessions and memory files without tight source-by-source consent.
Review this before installing if your OpenClaw sessions or memory files may contain private or work-sensitive information. Only approve lookups when you want prior-session content searched, and avoid approving broad recent-session searches unless you are comfortable with previews or history from unrelated conversations entering the current context.
SKILL.md:75Overbroad Cross-Session History and Shared-Memory Access
The skill is designed to search memory files and session history based on vague natural-language cues, then return matching content to the user. Because the trigger logic is broad and the retrieval/display path lacks strong authorization, source scoping, sensitivity filtering, or result minimization, it can expose private content from prior sessions or persistent memory that the user did not explicitly and knowingly request.
The activation examples include vague everyday phrases such as references to prior discussion that are common in normal conversation. This can cause the skill to trigger or prompt for cross-session retrieval when the user did not clearly request it, increasing the chance of unnecessary access to prior session data and unintended privacy exposure.
The detection scope is defined broadly across time references, platform references, content references, and behavior patterns, but it does not specify strong exclusion conditions or confidence thresholds. In a skill that can read memory files and session history, overbroad prompting materially raises the risk of social-engineering the user into approving retrieval of unrelated or sensitive historical context.
The natural-language guidance, prompts, and examples throughout the skill are presented exclusively in Chinese, which effectively forces a specific language for user interaction. The file does not state that the skill is region-specific or provide any user choice of language, which creates a locale/language policy concern.
The summary/description indicates the skill may trigger whenever a user mentions something not present in the current session, which is a broad conversational pattern and can cause the skill to activate in many normal interactions. Because the skill can read session history, search memory files, and write memory files, overbroad activation increases the chance of unnecessary cross-session memory access or prompting, creating privacy and consent risks even if explicit confirmation is later required.
The file’s comments and all generated user-facing messages are written in Chinese, including the inquiry and results display text. This forces a specific language experience without any visible opt-in, fallback, or documentation that the skill is intended only for a Chinese-language environment.
The skill initiates cross-session and memory-file searches without presenting a clear privacy warning at the point of retrieval or result display. Even though it asks whether to sync, it does not explain that this may access prior conversations or memory files containing sensitive data, which can lead to uninformed consent and unintended disclosure.
The package description is written entirely in Chinese and describes the skill behavior only in that language, which can impose a language/locale constraint without user opt-in. The file does not indicate that the skill is region-specific or that users may choose another language.
The summary and description are written only in Chinese, while the manifest does not state that the skill is region-specific or that language is user-selectable. This can indicate a language policy issue because the skill presentation appears to assume a specific language without opt-in or documented locale constraints.