Security audit
Security checks for vulnerabilities and agentic risk
The skill is mostly documentation with no executable code, but it requests broad write, network, and MCP authority while giving under-scoped prompts for installing skills and changing shared state.
Review this skill before installing in a tool-enabled agent. It should only be used in a trusted workspace where you are comfortable with possible file changes, external API use, MCP session/subagent activity, skill installation workflows, shared-state updates, archival logging, and scheduling actions. Require explicit approval before any write, network, installation, activation, or HQ state-change step.
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
---
name: "AI Company Framework"
slug: "ai-company-framework"
version: "3.0.0"
The trigger list contains very broad phrases like 'schema compliance', 'generalization', and 'naming conventions' that could match many ordinary requests and cause this skill to activate unexpectedly. Because the skill also declares write, network, and MCP capabilities, accidental invocation increases the chance of unnecessary file modification, outbound API use, or subagent/session actions beyond user intent.
The skill declares read/write file access, network API access, and MCP capabilities, but the user-facing description does not warn that using the skill may modify local files or communicate externally. This can mislead users about operational impact and reduce informed consent, especially in an agent ecosystem where skills may be auto-selected from metadata and triggers.
The instruction 'All content must be in English' imposes a specific language requirement in natural-language guidance. Under the policy, forcing a language without user choice or a clearly justified region-specific constraint is a violation.
The prompt explicitly directs logging execution metrics, updating shared state, archiving records, and scheduling follow-up actions without warning about data handling, authorization requirements, or side effects. In an environment where an AI agent may have connectors or automation tools, these instructions could trigger unauthorized persistence, state mutation, or disclosure of operational data.
The file presents a copy-paste prompt that instructs an AI to perform operational actions such as loading configuration, updating shared state via HQ, logging metrics, and archiving records, yet labels itself as 'not intended for automated agent invocation.' That mismatch can mislead users or downstream systems into treating the prompt as low risk while it still encourages system-affecting behavior if pasted into an agent with tool access.
The trigger keywords are extremely generic terms like 'Standard', 'Module', 'Pattern', and 'Register', which are likely to appear in ordinary user requests. In an agent skill system, this can cause unintended activation, routing confusion, or priority capture by this skill when another skill would be more appropriate, increasing the risk of incorrect behavior across the ecosystem.
The natural-language requirement "English-only for all compiled content" imposes a specific language choice across the skill framework, while only selectively allowing Chinese in trigger keywords. This is a language policy constraint presented without user choice, opt-in, or a clearly justified region-specific compliance reason.
No suspicious patterns detected.