Security audit
Security checks for vulnerabilities and agentic risk
This markdown-only skill gives process guidance for low-cost long-running GPT-5.4 workflows and does not contain hidden execution, exfiltration, or persistent system changes.
Install this only if you want an agent to prefer GPT-5.4 for long-running work and to create checkpoint artifacts for resumability. Review tasks involving external services, project trackers, or subagents so the agent's batching, writes, and delegation match your intended scope.
The skill explicitly directs use of a specific model ('GPT-5.4') as the default for orchestration and delegated work without requiring user opt-in. This can improperly constrain model selection, reduce operator control, and cause the agent to follow a hardcoded routing policy that may be suboptimal for safety, quality, or policy-compliance-sensitive tasks.
The file repeatedly frames usage as 'GPT-5.4 Only' and 'Use GPT-5.4-only mode,' which imposes a fixed model/language setting rather than offering a user choice. Under the policy, forcing a specific language or locale without opt-in is a natural-language policy concern; this is analogous in that it prescriptively constrains user-facing mode selection without documented justification.
No suspicious patterns detected.
