Security audit
Security checks for vulnerabilities and agentic risk
This JSON repair skill is not clearly malicious, but it asks for broad command execution while its instructions drift into unrelated API, network, and general automation behavior.
Review this skill before installing. It appears to be a local JSON repair helper, but its instructions are broader than necessary and could lead an agent to run shell commands, perform network troubleshooting, or treat the skill as useful for unrelated automation tasks. Use it only for explicit JSON repair on files you choose, and be aware that backup files may contain the same sensitive data as the originals.
The description uses broad activation language covering AI model calling, intelligent dialogue, agent orchestration, and LLM applications, far beyond JSON repair. This increases the chance the skill is selected for unrelated requests, after which its exec capability could be applied in contexts where local file repair was not intended.
The capability-scope keywords are excessively broad and generic, including terms for coding assistance, deployment, and fuzzy development use cases. In an agent ecosystem, such keywords can cause accidental routing to a skill with exec permissions, making unintended command execution more likely.
One section claims all repairs occur locally with no external services or data exfiltration risk, while nearby sections include assumptions and troubleshooting steps for network connectivity and external API-style operation. These contradictory assurances can mislead users and agents about the trust boundary, causing sensitive data to be handled under false assumptions and normalizing unnecessary network behavior.
The skill is presented as a lightweight local JSON repair tool, but later documentation expands into API integration, external service handling, and command/network-oriented capabilities that are not justified by the stated purpose. This scope drift can cause over-broad invocation and lead an agent to perform unintended actions beyond local file repair, increasing the attack surface and enabling misuse through ambiguous task matching.
The description states '支持中文交互', and the document consistently frames the skill around Chinese interaction without offering alternative language selection or opt-in. This can conflict with language or locale policy expectations when users have not chosen Chinese.
The feature note says the skill covers '中文交互' as a core capability, but does not clarify that other languages are supported or that the user's preference will be respected. This creates a natural-language policy issue if the skill defaults to one language without user opt-in.
Network troubleshooting guidance such as using ping and checking firewall/proxy settings is unjustified for a local JSON syntax repair skill. Even if not directly malicious, it can steer an agent into performing unrelated network diagnostics or environmental probing, which expands behavior beyond user intent and violates least privilege for this skill context.
No suspicious patterns detected.
