T06 · System Persistence
- Location
SKILL.md:53- Finding
Persistent Autonomous Agent Tasks Installed Through Gateway Cron
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
This OKR tracking skill is mostly coherent, but it can set up recurring autonomous agent jobs that keep reading and writing work data after the initial session.
Review the scheduled-report feature before installing. Only enable cron automation if you want recurring background agent sessions to process your workspace OKR data, and make sure you know how to list, pause, and delete those jobs. Avoid storing confidential personnel or performance data in synced or shared workspaces unless access is controlled.
SKILL.md:53Persistent Autonomous Agent Tasks Installed Through Gateway Cron
The skill explicitly says the agent will directly read and write JSON files in the workspace, but it does not clearly warn the user that this modifies persistent project data. Because the skill also frames this as 'zero dependency' and automatic tracking, users may enable it without understanding that ongoing file changes will occur, which can affect privacy, integrity, and auditability of workspace contents.
The cron setup instructions create isolated scheduled agent tasks that will continue to access workspace logs and generate reports after setup, yet the documentation does not clearly disclose this ongoing autonomous access. This is dangerous because users may not realize that background tasks will keep reading and writing sensitive work-history data on a recurring basis, increasing the chance of unintended data exposure or unreviewed modifications.
All user prompts, headings, and agent responses in this file are exclusively in Chinese, with no indication that the language is optional or region-specific. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation.
The guide explicitly instructs the agent to modify an existing date-based JSON work-log file in response to a natural-language request, but it does not require any confirmation, preview, backup, or audit trail before overwriting stored data. In a skill centered on persistent recordkeeping, silent in-place edits can cause accidental or unauthorized tampering of historical logs and reports derived from them.
The examples explicitly show the agent writing OKR configs and generated reports to local files, but provide no warning that potentially sensitive work logs, objectives, progress, and annual summaries will be persisted. In an OKR/work-tracking context, these files can contain confidential operational, personnel, and performance data, so silent persistence increases the risk of unintended disclosure on shared machines or synced directories.
All headings, prompts, and responses are written in Chinese, which implies a fixed language behavior. There is no indication that users may choose another language or that the skill is intentionally limited to a Chinese-language or region-specific context.
This markdown file demonstrates the agent saving data to local JSON files such as daily, plans, and weekly outputs, and later updating existing files. The examples do not include any warning that the skill persists user-provided work data to disk or that modification commands overwrite prior file contents.
该 markdown 文档从标题到正文均明确以中文编写,并直接面向“OKR 工作管理系统”的通用数据结构说明,但未说明这是仅限中文用户或特定地区场景的文档。按规则,未提供用户语言/locale 选择且无明确合理化说明的语言强制属于自然语言策略问题。
Automatically creating directories and files is a write operation that affects the user's workspace. In markdown guidance, such behavior should be disclosed clearly so users understand that invoking the skill may create persistent files on disk.