Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only workflow helper; its triggers are overly broad, but the artifacts do not show hidden execution, credential use, persistence, or data exfiltration.
Before installing, be aware that the skill may activate too easily because of generic keywords and implicit invocation. It appears safe from a security standpoint, but users may prefer narrower triggers so unrelated requests are not routed into this workflow helper.
The keyword list includes extremely common terms such as 'self' and 'improving', which can match ordinary user messages and unintentionally invoke the skill outside its intended scope. In an agent ecosystem, over-broad activation can cause prompt hijacking of normal workflows, unexpected execution paths, or accidental disclosure/manipulation of context by routing unrelated tasks into this skill.
The trigger phrases include extremely broad terms like "self", "improving", and "proactive," which are common in ordinary user requests and can cause the skill to activate outside its intended scope. In an agent setting, this can lead to inappropriate routing, accidental invocation, and unintended execution of workflow guidance in unrelated conversations, especially because the skill is positioned as proactive and self-improving.
The description says to use the skill when a user asks for broad concepts like "work-productivity," "self," "improving," or "proactive," which creates ambiguous activation criteria. In a skill-routing system, such broad matching increases the chance of misfires and unintended invocation, expanding attack surface and making it easier for adversarial prompts to steer execution into this skill.
The trigger list includes extremely generic terms such as "self," "improving," and "proactive," which are common in normal conversation and likely to cause accidental activation. This can route unrelated user requests into the skill, leading to inappropriate behavior, prompt/context hijacking opportunities, or unintended disclosure/use of workflow logic in contexts where it does not belong.
The trigger keywords include extremely broad everyday terms such as "self", "improving", and "proactive", which can cause the skill to activate for many unrelated requests. This creates routing confusion and may override or interfere with more appropriate skills, reducing predictability and potentially exposing users to unintended workflow changes or outputs.
The trigger sentence is framed in broad, ordinary language ('Help me ... practical help'), which can cause the skill to activate during unrelated conversations that merely resemble the phrasing. Unintended invocation is risky for an agentic workflow skill because it may steer the assistant into unsolicited planning or workflow execution when the user did not explicitly request this skill.
The example trigger sentence starts with a generic phrase like 'Help me', which is common in everyday conversation and does not enforce clear invocation boundaries. This increases the chance that unrelated user requests are interpreted as skill activation, causing untrusted content to be processed under this skill's instructions and potentially interfering with other agent behaviors.
This file presents the skill documentation entirely in Chinese while the skill content and audience description include broader, multilingual users, and there is no statement offering language selection or opt-in within this file. That creates a locale-policy concern because the user-facing documentation is fixed to a single language without an explicit choice.
The skill description defines applicability in very broad terms, covering many generic concepts like work productivity, self-improving, proactive behavior, workflows, artifacts, checklists, analysis, and implementation support. Ambiguous scope increases the chance of accidental invocation and inconsistent behavior, especially in environments with multiple overlapping skills.
The skill enables implicit invocation while using very broad, generic wording such as 'work-productivity', 'self-improving', and 'practical help' without clear scope boundaries. This can cause the agent to invoke the skill in unrelated contexts, increasing the chance of prompt-surface expansion, unintended behavior, or unsafe delegation to a loosely defined workflow helper.
The invocation guidance mixes keyword matching with ambiguous natural-language activation, without defining boundaries for when the skill should and should not run. In a productivity/self-improving agent context, that ambiguity increases the chance of accidental routing, overreach, or the skill hijacking generic requests containing common words like 'self,' 'improving,' or 'proactive.'
