Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only workflow helper with overly broad activation wording, but no hidden code, persistence, credential use, or destructive behavior.
Before installing, be aware that this skill may activate too easily for generic terms like self, logs, improvement, or bug fix. It appears safe from a security standpoint, but tighter trigger wording would reduce accidental use in unrelated conversations.
The trigger list is excessively broad, including generic terms like 'self', 'improving', and 'logs' that are common in ordinary conversations. This can cause the skill to activate unintentionally in unrelated contexts, creating prompt-routing confusion and increasing the chance that users are steered into this workflow when a different skill or no skill should apply.
The trigger keyword list includes extremely broad terms such as "self", "improving", "own", and "logs", which are common across many unrelated conversations. This can cause the skill to activate outside its intended scope, leading to inappropriate routing, confusing outputs, or accidental application of workflow guidance in contexts where it does not fit.
The trigger sentences and keywords are broad enough to match ordinary productivity requests such as 'help me', 'workflow', 'improvement', or 'bug fix', which can cause the skill to activate outside its intended niche. Over-broad activation increases the chance of context hijacking, inappropriate routing, and accidental invocation on unrelated user tasks, reducing user control and potentially exposing the agent to adversarial prompt overlap.
The trigger phrases are generic and match ordinary requests such as asking for help, needing a workflow, or fixing bugs. In a skill-routing system, this can cause the skill to activate for unrelated conversations, creating unintended prompt injection surface and allowing this skill to override or interfere with more appropriate skills.
The trigger keywords and example invocations are broad enough that the skill may activate for vague terms like 'self', 'logs', or 'improvement' outside the intended domain. In an agent ecosystem, this can cause unintended routing, overbroad application of the workflow, and processing of unrelated user requests, which increases the chance of unsafe or incorrect actions.
The manifest description uses expansive invocation guidance such as 'use when a user asks for work-productivity, self-improving-agent, self, improving, logs' without clear scope limits. This weakens routing precision and may cause the skill to be selected for loosely related requests, which can degrade system reliability and expose users to irrelevant or misleading workflow guidance.
The skill description says to use the skill for broad topics like work-productivity, self, improving, and logs without clearly defining boundaries or exclusions. Ambiguous activation criteria increase the chance of misclassification and over-broad invocation, which is especially risky for a helper intended to influence workflows and operational decisions.
The example trigger phrases are highly generic and reinforce activation on broad language rather than concrete task conditions. Without negative examples or tighter constraints, downstream systems or users may infer that nearly any bug-fixing or workflow request should invoke this skill.
The default prompt uses broad, common-language triggers like 'help me' and generic productivity terms, which can cause the skill to be selected in situations far beyond its intended scope. This increases the chance of unintended invocation, context confusion, and prompt-surface expansion where the skill influences unrelated tasks.
Enabling implicit invocation without tight activation constraints allows the platform to auto-select this skill based on vague overlap with ordinary user requests. In combination with the broad prompt/description, this can lead to accidental routing, over-collection of context, and execution of the wrong workflow in security- or reliability-sensitive situations.
Although this is a zh-CN README, core instructional content such as the title, requirement description, workflow description, keywords, and trigger phrases is written in English. This imposes a language choice on users without opt-in or explanation, which can conflict with language/locale policy expectations.
