Security audit
Security checks for vulnerabilities and agentic risk
This skill does not contain executable malware, but its automatic activation rules are broad enough that it could affect unrelated user requests.
Review this before installing if you run many skills or rely on implicit skill routing. The main risk is accidental activation during ordinary security, bug-fix, or workflow requests; tightening triggers to the exact skill name or disabling implicit invocation would make it more predictable.
The trigger phrases are broad, generic, and closely resemble normal user requests for help, which can cause the skill to activate in situations outside its intended scope. In an agent ecosystem, this increases the chance of unintended routing, priority hijacking, or the skill influencing unrelated tasks where security-sensitive workflow guidance may not be appropriate.
The description says to use the skill when a user asks for broad categories like 'work-productivity', 'skillscan', 'security', 'gate', or 'every', or whenever they need almost any workflow, checklist, analysis, or implementation support. This ambiguity greatly expands the activation surface, increasing the chance the skill is selected for unrelated tasks and creating prompt-routing confusion that can undermine least-privilege behavior.
The trigger list is extremely broad, including generic terms like 'security', 'every', 'must', 'pass', 'before', 'activate', and 'bug fix'. This makes unintended activation likely in normal conversations, which can cause the wrong skill to intercept unrelated requests and influence agent behavior in contexts where it was not intended.
The trigger list includes very common terms such as "security", "every", "must", "pass", and "before", which can match many unrelated user prompts and cause unintended skill activation. Overbroad activation increases the chance that this skill intercepts requests outside its intended scope, leading to confused task routing, policy bypass in multi-skill environments, or accidental execution of workflows the user did not request.
The default prompt includes a very broad natural-language trigger phrase covering generic terms like work-productivity, security, workflow, checklist, analysis, and implementation support. Combined with allow_implicit_invocation: true, this creates a realistic risk of unintended or silent skill activation during ordinary user requests, which can inject this skill's behavior into unrelated conversations and widen prompt-surface exposure.
The trigger examples include generic terms such as "security," "gate," and especially fragmented/common words like "every," "must," or "pass," which can cause the skill to activate in unrelated conversations. Over-broad invocation increases the chance of unintended routing, causing the wrong workflow to run in contexts involving sensitive security or operational decisions.
The trigger phrases are broad enough to match ordinary user requests about productivity, security, or bug fixing, which can cause the skill to activate outside the user's intent. In an agent ecosystem, this creates a prompt-routing or unsolicited activation risk where the skill may inject workflow behavior into unrelated conversations.
The keyword list is fragmented and excessively broad, including common terms like 'security', 'every', 'must', 'pass', and 'bug fix' that appear in many unrelated conversations. This makes accidental activation likely and can let the skill overreach into ordinary tasks, reducing predictability and potentially interfering with safer or more relevant skills.
The activation description says to use the skill when users mention broad categories like work-productivity, skillscan, security, or need almost any practical workflow, checklist, analysis, or implementation support for the requirement. This ambiguous scope can make the skill eligible for a very wide range of requests, increasing unintended activation and reducing isolation between skills in a shared agent environment.
The invocation guidance uses ambiguous natural-language examples and broad descriptions instead of precise activation criteria, making accidental invocation likely. In a security-oriented workflow helper, misrouting can lead users to receive irrelevant or misleading security process guidance when they did not intend to invoke this skill.
The file lists only English and Chinese instruction variants as the supported user-facing documentation, but does not state how language selection occurs or that the user may choose their preferred language. This can create an implicit locale constraint without documented opt-in.
