Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation/workflow helper with no code execution or persistence, though its activation wording is broad enough to cause accidental use in unrelated chats.
Before installing, be aware that this skill may be invoked too broadly because of generic trigger words and implicit invocation. It appears safe as a workflow helper, but users or maintainers should narrow the triggers to explicit skill-vetting or pre-install review requests to avoid confusing routing.
The manifest description includes very broad activation terms such as 'security' and 'first', which are common across many unrelated user requests. This can cause the skill to activate outside its intended scope, creating prompt-routing confusion and increasing the chance that users receive this workflow when a more appropriate or safer skill should handle the request.
The trigger keyword list contains vague terms like 'security', 'first', 'before', and 'github', all of which are frequently present in ordinary conversations unrelated to this skill. Overbroad keywords increase accidental invocation risk, which can hijack routing, expose users to irrelevant instructions, and reduce trust in the skill-selection system.
The trigger list includes very broad, high-frequency terms such as “security”, “first”, “before”, and “github”, which can cause this skill to activate in many unrelated conversations. Over-broad activation increases the chance of context hijacking or unintended routing, where users receive this workflow instead of a more appropriate skill, degrading safety and reliability of the agent system.
The trigger phrases are broad enough to match ordinary user requests containing generic terms like 'security', 'github', 'first', or 'before', which can cause this skill to activate outside its intended scope. In an agent-routing context, overbroad activation can hijack unrelated requests, leading to incorrect workflows, confusion, or unnecessary access to sensitive development or security-related tasks.
The activation guidance lacks precise boundaries and mixes broad keywords such as 'security', 'first', 'github', and 'bug fix' with open-ended trigger examples, making over-triggering likely. This is particularly dangerous in an agent ecosystem because ambiguous routing can cause this skill to preempt more appropriate tools or analyses, reducing reliability and potentially affecting security review quality.
The trigger sentence is so generic and fragmentary that it can match ordinary user requests unrelated to this skill’s intended scope, causing accidental activation. In a security- and vetting-themed skill, unintended invocation is risky because the agent may apply the wrong workflow, expose misleading analysis, or interfere with higher-priority safety/review flows.
The trigger phrases are overly broad and include generic help-seeking language, which can cause the skill to activate in unrelated conversations. In a security- and vetting-themed skill, accidental invocation is more dangerous because it may steer users into trusting or applying the workflow in contexts where it was not intended, increasing confusion, misrouting, or unsafe automation.
The trigger section provides positive examples but no clear activation boundaries, exclusions, or negative examples, so the system has little guidance on when not to invoke the skill. In context, this makes the broad keywords more dangerous because nothing in the trigger design counterbalances them, increasing false activations and misrouting during security-related or GitHub-related requests.
The description says to use the skill when users mention broad categories like work-productivity, security, or need any practical workflow, checklist, analysis, or implementation support, but it does not clearly bound the scope to skill-vetting scenarios. This ambiguity makes accidental invocation more likely, which can misroute user requests and cause the agent to apply the wrong workflow or assumptions.
The default prompt contains very broad activation language such as 'help me' plus generic productivity/security-oriented terms, which can overlap with ordinary user requests and cause unintended invocation. Because implicit invocation is enabled, this increases the chance the skill is triggered in unrelated conversations, exposing users to unexpected behavior and making prompt-routing less predictable.
This Chinese README contains substantial English instructional content and trigger phrases, but it does not state whether users may interact in Chinese, English, or choose their preferred language. That can amount to an implicit language constraint without user opt-in.
