Security audit
Security checks for vulnerabilities and agentic risk
This is a text-only skill, but its name says business migration while its instructions and triggers are for album filtering and are broad enough to activate unintentionally.
Review this skill before installing. It does not appear to run code or access sensitive data, but its packaging is internally inconsistent and its broad implicit triggers could cause it to be selected for unrelated requests. Install only after the publisher aligns the name, description, README, and trigger phrases to one narrow purpose.
The skill’s declared identity and its actual behavior do not match: it is named as a USA business migration planner, but the content is about filtering singles in an Albums view. This kind of mismatch can cause incorrect routing, user deception, and unsafe invocation in unrelated contexts because operators and downstream systems may trust the name/description rather than the body.
The trigger description includes extremely broad terms such as 'general-help' and generic workflow language, which can cause the skill to activate for many unrelated requests. In a skill-routing environment, this creates prompt-squatting risk where the wrong skill captures user intent and may override or interfere with more appropriate skills.
The keyword list uses vague single words like 'filter', 'out', 'singles', 'albums', and 'related', any of which may appear in ordinary conversation. This makes accidental or opportunistic activation much more likely, increasing the chance that the skill hijacks unrelated requests and injects off-topic instructions into the interaction.
The skill description defines invocation conditions in vague, expansive language, including generic request types like "general-help" and broad topic words. This ambiguity makes the routing boundary unclear and can cause the skill to be selected for unrelated conversations, which is risky because the skill's actual content is mismatched to its title and may steer users incorrectly.
The trigger keywords include extremely broad everyday terms such as "general-help", "filter", "out", and "related", which can cause the skill to activate for many unrelated user requests. Over-broad activation increases the chance that this skill intercepts prompts outside its intended domain, causing misrouting, confusing outputs, and unintended influence over agent behavior.
The skill metadata names the skill as a USA business migration planner, but the referenced requirement and trigger content are about filtering singles in an Albums view. This mismatch can cause the wrong skill to activate for unrelated requests, creating prompt-routing confusion and increasing the chance of unintended behavior or misuse of overly broad instructions.
The trigger phrase is written as ordinary conversational language and can easily match benign user requests that were not intended to invoke this skill. Overbroad activation is dangerous because it can hijack normal conversations, cause incorrect tool/skill selection, and amplify any harmful or irrelevant instructions packaged with the skill.
The skill is named 'usa-business-migration-planner', but the README describes functionality for filtering singles in an albums view, which is a materially different purpose. This mismatch can misroute user requests, confuse operators, and conceal a skill’s actual behavior, increasing the chance of unintended invocation or policy bypass through deceptive packaging.
The invocation section lacks clear constraints on when the skill should or should not activate, and it mixes generic keywords like 'general-help' with common words such as 'filter' and 'view'. In context, this makes the skill easier to trigger unintentionally and harder to govern safely, especially given the inconsistent stated purpose.
The trigger phrase is broad natural language that overlaps with ordinary requests for help, making accidental invocation more likely. Overbroad triggers can cause the wrong skill to activate for unrelated conversations, which is risky when the skill’s true scope is already ambiguous.
The trigger phrases are very broad and include common terms like 'general-help', 'filter', 'out', and 'related', which can cause the skill to activate in many unrelated conversations. Overbroad activation increases the chance of unintended invocation, context hijacking, and misrouting user requests to a skill that is not actually relevant.
The example trigger sentences are written in everyday, generic language and closely resemble natural user requests, which can train or bias routing systems toward over-matching. While less severe than the keyword list itself, this still broadens activation and reinforces prompt-squatting behavior.
The example trigger phrases overlap heavily with normal help-seeking language and provide no counterexamples showing when the skill should not run. This can train or bias routing toward selecting the skill too broadly, especially when combined with the vague description and broad keywords elsewhere in the file.
The skill enables implicit invocation without narrowly defined trigger constraints, while its metadata and prompt are broad and internally inconsistent. This can cause the agent to invoke the skill in contexts the user did not clearly intend, increasing the chance of irrelevant actions, misrouting, or unauthorized influence over user workflows. The odd mismatch between the skill name ('USA Business Migration Planner') and the described task ('Filter out singles in Albums view') makes accidental invocation more likely, which raises risk in practice.
The activation description uses generic terms like general-help and common words such as filter, out, singles, albums, and related, without clear boundaries on when the skill should run. This ambiguity increases the attack surface for prompt injection via routing, because many unrelated user inputs could satisfy the activation conditions and invoke a mismatched skill.
This is a Chinese-language README, but the documented trigger phrases are presented only in English and imply that activation depends on English wording. That can amount to a language-policy issue when no user language choice or justification is provided.
The skill is named and described as USA Business Migration Planner, which imposes a specific locale context in the natural-language metadata. There is no indication that users can opt into a different locale or that the U.S.-only scope is explicitly justified in the file.
