Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only workflow helper, but its activation wording is too broad and may cause it to appear in unrelated security or productivity requests.
Installers should be aware that the skill may be selected too often because of generic trigger terms and implicit invocation. It is otherwise a low-impact documentation/workflow helper with no code execution or persistence, so the main improvement would be narrowing activation to explicit SkillScan-style workflow requests.
The trigger keyword list is overly broad and includes generic terms such as "security," "every," "must," "pass," and "before," which can cause the skill to activate in many unrelated conversations. This increases the chance of unintended invocation, which can confuse routing, override more appropriate skills, and expand the attack surface for prompt/skill selection abuse.
The skill description says to use the skill when a user asks for broad terms like 'work-productivity', 'security', or 'artifact, checklist, analysis, or implementation support', which creates an extremely large activation surface. In an agentic system, this can cause unintended routing or invocation on unrelated requests, potentially exposing users to the wrong workflow, suppressing more appropriate safeguards, or causing actions to be taken under an incorrect skill context.
The keyword list includes generic everyday words such as 'security', 'gate', 'every', 'must', 'pass', 'before', and 'activate', many of which appear in normal conversation with no relation to this skill. This makes accidental invocation highly likely and can let the skill preempt more suitable tools or analyses, which is especially risky for a security-themed helper that may influence decisions about safety gates and bug fixes.
The trigger keyword list includes extremely broad everyday terms such as "security", "every", "must", "pass", and "before", which can cause the skill to activate in many unrelated conversations. Over-broad activation can unexpectedly inject this skill's workflow into other tasks, creating prompt-scope confusion and potentially overriding more appropriate skills or user intent.
The trigger keywords are excessively broad and include generic terms like "security," "every," "must," and "before," which can cause the skill to activate in many unrelated conversations. In an agent environment, unintended activation can route sensitive or unrelated tasks into this skill, increasing the chance of incorrect behavior, workflow hijacking, or accidental exposure of user context to an inappropriate prompt path.
The example trigger sentences are broad natural-language requests that match common help-seeking language rather than clearly scoped invocations. This makes accidental triggering more likely, especially because the examples overlap with ordinary requests about bug fixing or hardening, which can cause the agent to invoke the skill outside its intended context.
The example trigger phrases are broad, awkward, and not well-scoped to a clear user intent, making it unclear when the skill should activate. Ambiguous activation examples can lead to accidental routing on partial phrase matches or templated prompts, reducing reliability and potentially causing the wrong skill to handle sensitive security-related requests.
The example trigger sentences are built around highly generic openings like 'Help me' and 'I need', combined with pasted requirement text, which signals that ordinary user requests could match the skill too easily. While examples are indirect compared with explicit trigger keywords, they still reinforce unsafe activation patterns and may be consumed by systems or authors as de facto invocation guidance.
该技能文件为 SKILL.zh-CN.md，整体内容固定为简体中文环境，但文中未说明用户可切换语言，也未提供语言偏好选择或 opt-in 机制。对于面向广泛用户群的技能，这可能构成语言/区域策略上的自然语言约束。
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 a practical workflow or checklist for the requirement. This boundary is vague enough to match many benign requests, increasing the risk of unintended activation and misleading task routing.
The skill combines a vague, marketing-style default prompt with allow_implicit_invocation enabled, which can cause the agent to activate this skill in situations only loosely related to the user’s request. That broad trigger surface increases the chance of unintended invocation, prompt-context pollution, and unsafe workflow interference, especially for a skill framed around security and implementation support.
The trigger phrase is broad enough to match ordinary user language rather than a narrowly scoped invocation, which can cause unintended skill activation. In a security-oriented workflow, accidental activation is risky because it may route unrelated requests through this skill, creating confusion, overreach, or unsafe automation in contexts the user did not intend.
Although the sentence includes more context, it still reads like generic help-seeking language and lacks a precise boundary that distinguishes when this skill should activate. This increases the chance of misrouting benign conversations into a security workflow, which is especially problematic for a skill associated with scanning, gating, and implementation guidance.
