Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only ontology workflow skill with some overbroad localized/reference wording but no hidden execution, data access, persistence, or destructive behavior.
Installing this skill is reasonable if you want help designing typed ontologies or lightweight knowledge graphs. Be aware that the Chinese localization and requirement-plan reference are broader than the main English skill, so users should invoke it explicitly for ontology, schema, entity-relation, validation, migration, or query-pattern tasks rather than generic bug-fixing work.
The trigger sentence begins with a very broad everyday help-request pattern, making accidental invocation likely. In agent systems, overbroad activation logic can route unrelated user requests into this skill, causing prompt-scope confusion, unreliable behavior, and possible policy bypass if the skill's instructions conflict with the correct specialized handler.
This markdown file is entirely written in Simplified Chinese and does not indicate that other languages are available or that Chinese is optional. Under the policy criteria, forcing a specific language without user opt-in can be a natural-language policy violation unless the locale restriction is clearly documented and justified.
The manifest and body describe a generic demand-validation and workflow-helping skill rather than the declared ontology/knowledge-graph design helper. This kind of semantic mismatch can cause the wrong skill to be selected and trusted for tasks outside its stated scope, increasing the chance of unsafe automation, incorrect outputs, or hidden behavior under an unrelated label.
The top-level description is broad enough to match many common requests, including generic implementation support, analysis, or workflow help. In an agent environment, such broad matching increases unintended activation and can cause the model to apply the wrong instructions or tools to user requests.
The trigger keywords and example invocations are framed around broad productivity and bug-fixing help instead of ontology/schema/graph design. Overbroad and misaligned triggers make accidental invocation likely, which can route unrelated requests into this skill and produce irrelevant or potentially unsafe actions under false assumptions.
The keyword list includes generic terms like 'structured', 'memory', 'creating', and 'bug fix', which overlap heavily with ordinary user requests unrelated to ontology design. In skill-routing systems, this can hijack benign conversations and inject unrelated workflow instructions, degrading reliability and potentially steering actions incorrectly.
The surrounding skill manifest says this skill is for designing typed ontology and knowledge-graph workflows, but this file's 'Live Requirement' and 'Audience' describe generalized help with fixing bugs, hardening setup and safety, improving reliability, and creating adjacent skills. That is not merely incomplete documentation; it actively reframes the skill's purpose away from ontology/schema design into broad skill-improvement work.
The 'How The Skill Meets The Requirement' and 'Executable Implementation Plan' sections describe a generic process that can produce templates, automation outlines, checklists, code changes, or decision support. This contradicts the manifest's claimed specialized function of designing typed ontologies and knowledge-graph workflows, because the documentation presents a materially broader operational intent.
This trigger remains ambiguous about when the skill should activate and does not clearly bind activation to ontology or knowledge-graph tasks. In a multi-skill agent environment, ambiguous triggers increase false positives in routing and can degrade safety by sending requests to a skill that was not designed for the actual task.
The usage signals and trigger phrases advertise broad bug-fixing and hardening help instead of narrowly scoped ontology workflow support. In an agent-routing context, overly broad skill descriptions can cause unintended activation, pulling the agent into handling tasks outside its validated scope and increasing the chance of unsafe or misleading outputs.
