Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only workflow helper with sloppy, overly broad activation wording but no executable code, persistence, credential access, or data-exfiltration behavior.
This skill appears safe to install from a security perspective, but users should be aware that its generic trigger words and implicit invocation setting may make it activate for unrelated productivity or phone-number-related prompts. Prefer explicit invocation by skill name if the platform allows it.
The documented trigger phrases are broad natural-language requests that can match ordinary user prompts unrelated to this skill’s intended scope. In an agent ecosystem, this can cause accidental invocation, prompt-routing confusion, and unintended execution of the skill in contexts where the user did not explicitly request it, increasing the chance of unsafe or incorrect automation.
The description says to use the skill whenever a user asks for broad categories like 'work-productivity' or 'needs a practical workflow, artifact, checklist, analysis, or implementation support,' which is so expansive that it blurs the intended boundary of the skill. In an agentic environment, ambiguous invocation criteria can lead to over-selection, causing this skill to intercept unrelated tasks and provide mismatched guidance.
The trigger list includes extremely generic terms like 'every', 'phone', 'number', 'ability', 'things', and 'done', which are likely to match many unrelated user requests. That can cause accidental invocation of this skill in contexts it was not intended for, increasing prompt-routing errors and the chance that irrelevant or unsafe workflow guidance is injected into unrelated tasks.
The trigger keywords include extremely generic everyday terms such as “every”, “phone”, “number”, “ability”, “things”, and “done”, which makes accidental activation very likely. In an agent environment, over-broad invocation can route unrelated user requests into this skill, causing unintended workflow execution, context confusion, and potentially unsafe or irrelevant guidance.
The trigger text is so broad and everyday-language-like that it can cause accidental or inappropriate activation outside the intended PollyReach workflow context. In an agent environment, overbroad invocation patterns can route unrelated user requests into the wrong skill, causing misexecution, confusing outputs, and possible chaining into higher-risk actions if downstream workflows assume relevance.
The keyword list includes highly generic terms such as 'every', 'phone', 'number', 'ability', 'things', and 'done', which are common in everyday prompts. Such unspecific activation guidance materially raises the risk that unrelated user inputs will spuriously match this skill, leading to misrouting and unpredictable agent behavior.
The trigger examples are broad, natural-language phrases that could match ordinary user requests unrelated to this specific skill. That increases the chance of unintended skill invocation, which can cause the agent to apply the wrong workflow, produce irrelevant actions or artifacts, and create downstream safety or reliability issues.
This zh-CN README presents substantial user-facing descriptive content in English, despite being labeled as the Chinese version. That can amount to an implicit language policy choice for users of this locale-specific document without opt-in or explanation.
The keyword list includes highly generic words such as 'every', 'phone', 'number', 'ability', 'things', and 'done', which are common in everyday conversation and not specific to the skill. This broad scope can cause over-triggering and misrouting, making it easier for unrelated conversations to activate the skill and potentially expose users to unintended workflow actions or misleading outputs.
The example triggers are malformed, truncated, and largely repeat vague wording instead of showing realistic, specific activation conditions. Poor examples make routing behavior harder to validate and increase the risk that maintainers or automated systems will activate the skill based on ambiguous or low-signal phrases.
该文件为 SKILL.zh-CN.md，全文以中文提供说明，但未见任何关于用户可选择其他语言或根据用户偏好切换语言的说明。按规则，若技能在自然语言层面默认强制特定语言而无用户选择，可能构成语言/区域策略违规。
The skill description says it should be used when a user mentions broad terms or needs practical workflow/checklist/analysis support for the requirement, but it does not sharply constrain when the skill should or should not activate. That ambiguity increases the chance that unrelated productivity or implementation requests are captured by this skill, leading to incorrect handling and unnecessary exposure of skill-specific behavior.
The skill enables allow_implicit_invocation without defining any activation constraints, trigger phrases, or narrowing conditions. This can cause the agent to invoke the skill in situations the user did not clearly intend, which increases the chance of prompt-context hijacking, unexpected actions, or disclosure of irrelevant or sensitive workflow guidance.
This trigger sentence is insufficiently specific about when the skill should run, making activation criteria ambiguous. Ambiguous routing in agent systems is dangerous because it increases the chance that the skill is selected for unrelated prompts, which can bypass user intent, reduce reliability, and create unsafe workflow composition when multiple skills compete for similar broad requests.
No suspicious patterns detected.
