Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only workflow helper with overly broad trigger wording, but no hidden code, persistence, credential handling, or destructive behavior was found.
Install only if you want a general PollyReach-style workflow helper. Be aware that its trigger terms are too broad, so it could activate on unrelated prompts unless invocation is narrowed or implicit invocation is disabled.
The trigger keywords and example invocations are extremely broad, including common words like "every," "phone," "number," and generic workflow requests. This can cause the skill to activate on unrelated user prompts, leading to unintended prompt injection surface expansion, misrouting, or inappropriate handling of benign requests under this skill’s instructions.
The invocation description says to use the skill when a user asks for broad concepts like work-productivity or practical workflow help, which are common requests shared by many unrelated skills. Such ambiguous activation criteria increase the chance of accidental routing and make it easier for this skill to overmatch benign user requests outside its intended scope.
The trigger keyword list contains highly generic terms such as "every," "phone," "number," "ability," "things," and "done," which are likely to appear in unrelated conversations. This creates a broad, unsafe activation surface where the skill may be invoked unintentionally, causing routing errors, context hijacking, or inappropriate application of the skill in unrelated tasks.
The trigger keyword list includes highly generic everyday terms such as “every”, “phone”, “number”, “ability”, “things”, and “done”, which can cause the skill to activate in many unrelated conversations. Over-broad invocation expands the skill’s execution surface, increasing the chance of unintended routing, confusing outputs, and accidental interference with other more appropriate skills.
The trigger sentence includes broad, everyday terms such as 'every', 'phone', and 'number' in a skill intended for PollyReach-style workflow assistance. This creates a realistic risk of accidental invocation in unrelated conversations, causing the wrong skill to activate and potentially produce irrelevant or unsafe actions under false context. The surrounding file shows loose trigger construction rather than deliberate abuse, so the issue appears to stem from poor scoping rather than malicious intent.
The keyword list includes very common terms such as 'every', 'phone', 'number', 'ability', and 'things', which are likely to appear in unrelated conversations. This creates a broad activation surface that can cause unintended skill selection and potentially expose the user to irrelevant instructions or interfere with safer, more appropriate routing.
The trigger sentences are highly generic and embed broad natural-language fragments like 'help me' and 'I need a practical workflow', which can match many unrelated user requests. In an agent ecosystem, this increases the chance of accidental invocation, context hijacking, or routing a user into this skill when they did not intend to use it.
The invocation guidance describes when to use the skill in very general terms but does not define clear boundaries or exclusions. In agent systems, ambiguous routing criteria increase the chance that the skill is selected for unrelated tasks, which can create unsafe behavior, incorrect outputs, or unexpected access to workflow logic outside its intended scope.
The document is a Chinese-language README and references separate English and Chinese files, but it does not state that the user may choose their preferred language or locale. This can amount to a locale policy issue when the skill documentation defaults to a fixed language experience without explicit opt-in or selection guidance.
The example trigger phrases are vague and generic, demonstrating activation on broad requests like "I need a practical workflow," which does not sufficiently distinguish this skill from many others. Examples shape downstream matching behavior, so unsafe examples can reinforce overbroad invocation and increase false activations.
The description says the skill should be used when users mention broad terms like work-productivity, pollyreach, gives, every, or phone, or whenever they need a practical workflow, artifact, checklist, analysis, or implementation support for the requirement. This boundary is too vague and functionally allows activation for a wide class of routine requests, making accidental triggering and misrouting materially more likely.
The example trigger phrases begin with generic prompts like “Help me” and “I need a practical workflow for,” which are common across many benign requests. While examples alone do not force activation, they reinforce permissive matching behavior and weaken trigger specificity, which can contribute to over-selection of this skill in unrelated contexts.
The default prompt is overly broad and uses generic wording that can trigger in routine conversations unrelated to the intended PollyReach workflow. Combined with implicit invocation, this increases the chance the skill activates unexpectedly, causing prompt/context injection into unrelated tasks and potentially influencing agent behavior without clear user intent.
The trigger description and examples do not clearly define when the skill should activate, and they include vague, high-frequency language unrelated to a precise security or productivity intent. In an agent ecosystem, ambiguous activation criteria increase the chance of unintended routing, which can expose users to incorrect automation, context leakage into the wrong skill, or unreliable task execution. The work-productivity context makes this somewhat less severe than a privileged execution skill, but still operationally dangerous because misfires are likely.
No suspicious patterns detected.
