Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-style workflow helper with overly broad activation wording, but no hidden code, persistence, credential handling, or unsafe execution behavior.
Installers should expect a general workflow-planning helper for Nano PDF-style tasks. Consider narrowing or disabling implicit activation if you do not want ordinary PDF, editing, or general productivity requests routed into this skill.
The trigger phrases are broad enough to match routine requests involving work productivity, PDFs, editing, or generic workflow help, which can cause the skill to activate outside its intended scope. In an agent environment, this increases the chance of unintended invocation, prompt hijacking of normal user tasks, and routing sensitive document-related requests into an unrelated skill path.
The top-level description says to use the skill whenever a user asks for broad categories like "work-productivity," "pdf," "edit," or any practical workflow, artifact, checklist, analysis, or implementation support for the requirement. That scope is excessively broad and ambiguous, which increases the chance of accidental invocation and cross-domain capture of requests that do not actually belong to this skill.
The trigger list includes very generic terms such as "nano," "pdf," "edit," "natural," "language," and "instructions," which are common across many unrelated requests. This can cause the skill to activate outside its intended scope, leading to misrouting, unexpected behavior, or inappropriate application of workflow guidance in contexts where a different skill or stricter handling is needed.
The trigger list includes very generic terms such as "nano", "pdf", "edit", "natural", and "language", which are common across many unrelated requests. This can cause accidental skill activation, routing user prompts into an irrelevant workflow and increasing the chance of inappropriate instructions, confusion, or unsafe task handling in contexts the skill was not designed for.
The trigger sentence is broad enough to match ordinary user phrasing such as generic requests for help with workflows, PDFs, or editing, which can cause the skill to activate outside its intended scope. Over-broad activation increases the chance of prompt hijacking, unintended routing, or inappropriate use of this skill in contexts where its assumptions and outputs do not fit.
The activation guidance lacks boundaries, exclusion criteria, and disambiguation rules, which makes accidental invocation more likely across common user requests involving PDFs, editing, or productivity. In this context, the skill is not directly performing a dangerous domain action, but ambiguous routing still increases the chance of unintended behavior, misuse of agent context, and interference with more appropriate skills.
The trigger examples are phrased as generic help-seeking requests rather than narrowly scoped activation conditions, so the skill may activate in unrelated conversations about PDFs, editing, or general workflow help. In an agent ecosystem, over-broad invocation can cause unintended routing, prompt shadowing, or execution of the wrong skill, reducing reliability and potentially exposing downstream tools or data to an inapplicable workflow.
The file is presented as a Chinese README, but multiple core user-facing sections, including the title, requirement description, workflow description, and trigger phrases, are written in English. This can amount to a language/locale policy issue because the skill does not offer the user a language option or explain why English is being forced in a zh-CN document.
The invocation description says to use the skill for broad categories like work-productivity, nano, pdf, or practical workflow support, without tightly scoping the intended task boundary. This ambiguity makes over-selection more likely, which can misroute user requests and reduce reliability by applying this skill in situations where another skill or general handling would be safer and more appropriate.
The default prompt embeds a very broad natural-language activation phrase ('help me' plus generic productivity/PDF wording) that can overlap with ordinary user requests, increasing the chance of unintended invocation. Because implicit invocation is enabled, normal conversation about work productivity or PDFs could trigger this skill unexpectedly and route user intent into behavior they did not explicitly request.
The invocation guidance lacks precise activation criteria and instead uses vague, high-level language covering common tasks like practical workflow help, analysis, or implementation support. In an agent ecosystem, unclear boundaries can cause accidental invocation collisions with other skills, leading to incorrect task routing, context confusion, and potentially unsafe handling of unrelated user requests.
