Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-style workflow helper with overly broad activation wording, but no hidden execution, credential access, persistence, or destructive behavior.
Install only if you want a workflow helper for agent reliability, bug fixing, and self-improving workflow planning. Expect possible over-triggering because its keywords are broad; prefer explicit invocation by skill name when you want it used.
The manifest description says to use the skill when a user asks for broad topics like 'work-productivity', 'self', 'improving', or 'logs', which creates an ambiguous activation boundary. This can cause the system to select the skill for unrelated requests, expanding its operational scope beyond intended use and increasing the chance of incorrect, confusing, or unsafe assistance.
The trigger list includes extremely generic terms such as 'self', 'improving', 'logs', 'own', and 'findings', which are common in normal conversations and likely to cause unintended activation. Over-broad activation can route unrelated user requests into this skill, leading to inappropriate guidance, reduced reliability, and possible security issues if the skill is invoked in contexts it was not designed to handle.
The trigger list includes very generic terms such as "self", "improving", "logs", and "own", which can match many unrelated user requests. This can cause the skill to activate unexpectedly, overriding more appropriate skills or injecting irrelevant workflow guidance into normal conversations, reducing reliability and potentially causing unsafe task routing.
The default prompt embeds a long, generic trigger phrase including common terms like 'help me', 'work-productivity', and 'self-improving', which increases the chance of accidental or implicit invocation during ordinary user conversations. Because implicit invocation is enabled, this broad matching can cause the skill to activate unexpectedly, exposing users to unintended prompt injection surface, workflow changes, or unintended actions driven by the skill.
The trigger sentences are extremely broad and include common phrases like 'help me' and 'I need a practical workflow', which can cause this skill to activate for many unrelated requests. In an agent ecosystem, overly broad routing increases the chance of incorrect skill selection, prompt-context contamination, and execution of workflows outside the user's actual intent.
The keyword list includes extremely broad terms like 'self', 'improving', 'logs', 'own', and 'findings', which are common in unrelated conversations and can lead to accidental activation. Because this skill is framed around self-improving workflows, false activations could redirect user sessions, expose internal context to the skill unnecessarily, or cause unintended workflow guidance to be applied in unrelated tasks.
The trigger examples are written as natural, everyday requests rather than narrowly scoped activation phrases, which can cause the skill to fire when a user is making a general request instead of intentionally invoking this specific workflow. In a self-improving agent context, unintended invocation is more concerning because the skill may steer execution, analysis, or workflow changes in situations where the user did not explicitly request this behavior.
The trigger phrases are broad and include generic terms like "self," "improving," and "logs," which can cause the skill to activate in many unrelated contexts. In an agent ecosystem, overly permissive activation boundaries increase the chance of unintended invocation, context capture, and workflow execution on requests the user did not actually intend for this skill.
The skill description defines activation conditions broadly across multiple vague concepts and request types, without clearly stating when the skill should not be used. That ambiguity increases the chance of over-triggering, which can misroute user requests and lead the agent to apply the wrong workflow or produce irrelevant changes in contexts not actually related to self-improving agent workflows.
The trigger examples only show positive matches and do not provide contrasting negative examples or constraints. Without boundaries, implementers and routers are more likely to match broad natural-language requests that merely resemble the topic, which increases accidental invocation and degrades precision.
The keyword list contains highly generic terms such as 'self', 'improving', 'logs', 'own', and 'improved', which are likely to appear in ordinary conversation unrelated to this skill. Ambiguous trigger keywords can cause accidental invocation, leading to misrouting, unintended exposure of user context to the wrong skill, and reduced trust in the agent's behavior.
The file is a Chinese README, but core descriptive content and trigger text are partially written in English without stating that users can choose their preferred language. This can impose a language assumption on users rather than offering an explicit language or locale opt-in.
