Security audit
Security checks for vulnerabilities and agentic risk
This is a markdown-only goal-planning skill whose stateful weekly-planning behavior is disclosed in the artifacts and aligned with its purpose, though users should understand that it may use saved goal context when provided.
Before installing, understand that this skill may use saved goal, task, schedule, and progress context if the host app provides it, and it may help create weekly plans or reviews. It does not contain code or external network behavior, but users should keep control over what planning context is saved, reviewed, or cleared.
The description says to use the skill when a user has a 'vague, oversized, or tangled goal and wants help thinking it through,' which is a broad, subjective trigger without clear boundaries or exclusion conditions. This can overlap with many common planning or advice-seeking conversations and may cause unintended invocation.
The file is explicitly framed as an "English Guide" and does not offer users a language choice or explain a justified region-specific requirement. This can violate language/locale policy when the skill behavior is constrained to English by default rather than by user preference.
The 'when to use' criteria are broad enough to match many ordinary planning or help-seeking requests, which raises the risk of accidental invocation. Unintended routing can lead users into a multi-turn probing workflow they did not request, potentially eliciting unnecessary personal or operational details and displacing a more appropriate skill.
The guide materially expands a simple conversational clarification skill into progress tracking, weekly cycle reviews, schedule generation, and JSON/system-integration behavior. That broadening can cause the skill to collect and act on more contextual state than users expect, increasing the chance of unintended autonomy, cross-feature coupling, and misuse of background context in ways that exceed the original scope.
The dedicated trigger list uses vague, high-frequency phrases like 'help me think this through' and 'help me make a plan' that commonly appear in unrelated conversations. Without disambiguation logic, these triggers increase collision risk and can activate the skill in contexts where reflective questioning, background-context use, or planning behavior is inappropriate.
标题明确将该技能定义为“中文指南”，全文也未提供语言切换、用户选择或适用范围说明。若该技能在通用环境中被调用，可能构成未经用户选择的语言/locale 限制。
“想做一件事，但不知道从哪里开始”“需要一个计划”“有很多方向，不知道先做哪个”等描述虽然与技能目的相关，但仍然较为宽泛，且“常见表达”中包含多条通用求助语句。文档虽提供了不适用场景，但没有给出足够具体的触发范围定义或负例来防止与一般咨询/聊天请求重叠。
The guide expands a conversational goal-clarification skill into workflow orchestration by instructing it to emit integration-oriented scheduling artifacts such as a weeklyPlan JSON block. That creates a hidden side-channel from conversation into downstream systems and can cause unintended automation, data propagation, or action execution beyond what the user explicitly requested.
该段将“帮我理一理”“帮我做个计划”“我有很多想法”等常见日常表达列为触发短语，但没有限定具体领域、上下文或排除条件。这类短语在大量普通对话中都可能出现，容易导致技能被意外触发。
The skill is authorized to consume hidden [GOAL_CONTEXT] and [WEEKLY_CYCLE_REVIEW] system data and use it to steer responses without surfacing that dependency to the user. This violates least-privilege expectations for a clarification skill and risks privacy leakage, overcollection, and behavior drift into progress surveillance or workflow management.
The file title identifies this as an English workflow ('workflow-en'), which signals a language-specific behavior. The document does not state that users can opt into another language or that the English-only constraint is required for a region-specific or compliance reason.
The workflow instructions, prompts, and required outputs are all specified in Chinese, which effectively forces a specific language for interaction. The file does not indicate that Chinese is optional, user-selected, or required for a region-specific use case.
The workflow materially expands from conversational goal clarification into structured weekly scheduling, calendar generation, and ongoing review automation. That scope creep can cause the agent to perform quasi-project-management functions not disclosed by the skill's stated purpose, increasing the chance of over-collection of user data, persistent behavioral tracking, and user reliance on automated planning decisions.
Automatic trigger-based weekly cycle behavior introduces agent actions that may occur without a fresh, explicit user request, which is beyond the expected interaction model for a clarification skill. In context, this is risky because it encourages unsolicited workflow management and could surprise users with stateful follow-up behavior they did not knowingly authorize.
This markdown file contains user-facing natural-language content only in Chinese, which can be interpreted as forcing a specific language/locale without offering the user a choice. The policy allows locale constraints when they are clearly documented and justified, but this file does not state that it is intentionally Chinese-only or region-specific.
Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.
This is not only about building one skill. It is also a way to validate a bigger direction: whether high-quality clarification + action planning can become a reusable agent capability.
# Current Reality
You have strong interest, strong context, and a real use case. The main constraints are limited weekly time and a tendency to expand scope.
# Recommended Focus
Do not treat this as a full product yet. The best current focus is to build a stable self-use version that you would genuinely use repeatedly.
该文件标题和全部示例内容均以中文编写，并以中文助手回复作为默认呈现方式，但没有说明这是面向特定中文场景，或允许用户选择其他语言。按照语言/地区政策，若技能默认强制单一语言而无用户选择或明确合理限制，属于自然语言层面的策略风险。
The workflow relies on persistent contextual state such as GOAL_CONTEXT, ACTIVE_PHASE, and completion-rate analysis to track progress over time, which exceeds a simple conversational clarification role. While not inherently malicious, this creates unnecessary state retention and profiling risk if the system stores behavioral patterns, missed tasks, or inferred productivity traits without clear minimization and disclosure.
No suspicious patterns detected.