Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only workflow helper with no executable or persistent behavior, but its trigger wording is too broad and may activate unintentionally.
Before installing, be aware that this skill may be invoked for unrelated requests because its trigger keywords are very generic and implicit invocation is enabled. It appears safe from a code-execution and persistence perspective, but users may want the publisher to narrow the triggers to PollyReach-style workflow requests.
The trigger phrases are extremely broad and include common terms like 'help me', 'practical workflow', 'phone', and 'done', which can cause the skill to activate for unrelated everyday requests. In an agent ecosystem, this creates unintended routing and prompt capture risk, where this skill may intercept tasks outside its intended scope and influence outputs unexpectedly.
The trigger keywords and example invocations are overly broad, including generic terms like "every," "phone," "number," "ability," and "things," which can cause the skill to activate for unrelated user requests. In an agent ecosystem, this creates prompt-routing confusion and unintended invocation risk, which can divert tasks into the wrong workflow and potentially expose user context to an unintended skill.
The skill description is framed so broadly that it can match many generic productivity requests rather than a narrowly scoped PollyReach-related use case. This increases the chance of unintended invocation, causing the wrong skill to activate, override a more appropriate skill, or process user requests outside its intended domain.
The trigger keyword list includes very common words such as 'every', 'phone', 'number', 'ability', 'things', and 'done', which are likely to appear in unrelated conversations. In an agent environment, such low-specificity triggers can cause frequent accidental activation and create routing confusion or unintended tool use.
The trigger keyword list includes extremely broad, high-frequency terms such as "every", "phone", "number", "ability", "things", and "done". This can cause the skill to activate on unrelated user requests, leading to incorrect routing, prompt-context pollution, and possible overshadowing of more appropriate skills; in agent systems, that increases the chance that users receive irrelevant or unsafe automation guidance in the wrong context.
The default prompt is highly generic and includes broad everyday terms like 'help me' and common workflow/productivity language, which can cause the skill to be invoked in unrelated conversations. Because implicit invocation is enabled, this increases the chance of unintentional routing into this skill, expanding the prompt-injection and mis-execution surface even when the user did not explicitly request it.
The trigger content includes broad everyday terms such as 'phone', 'number', and 'things done', which can cause the skill to activate in many unrelated conversations. Overly generic activation language increases the chance of unintended invocation, causing context hijacking or routing users into this workflow when they asked for something else.
This file is named as a Chinese README, but substantial user-facing content is presented in English, which can violate language/locale expectations for users selecting the zh-CN version. The document does not offer a language choice or explain that the localized file intentionally contains English text.
The activation examples presented to users in the Chinese README are entirely in English, which imposes a language constraint on invocation despite the locale-specific document context. There is no explicit user opt-in, alternative Chinese trigger phrasing, or documented reason for requiring English.
The example trigger phrases are vague, truncated, and do not clearly define when the skill should or should not activate. This weakens operator and system understanding of the skill boundary, making misrouting more likely and reducing confidence that the skill is invoked only for appropriate tasks.
The skill description says it should be used when users mention broad terms like work-productivity, pollyreach, gives, every, or phone, or when they need general workflow or implementation support. These boundaries are too vague, so the skill may match a wide range of unrelated requests, increasing accidental invocation and making downstream agent behavior less predictable and less safe.
The example trigger phrases are themselves broad and templated, encouraging activation on generic requests like needing a practical workflow. Because examples often shape routing behavior and author expectations, this amplifies the overbroad-trigger problem and makes unintended invocation more likely across normal conversations.
The skill description describes a wide class of user needs ('practical workflow, artifact, checklist, analysis, or implementation support') without clear boundaries for when the skill should activate. In combination with allow_implicit_invocation: true, this makes accidental invocation more likely and weakens control over which user requests are routed into the skill.
The invocation examples are vague and do not define when the skill should or should not be used, leaving trigger scope ambiguous. In an agent environment, ambiguous routing rules can misfire across unrelated tasks, degrading reliability and potentially exposing user context to an unnecessary workflow path.
No suspicious patterns detected.
