Security audit
Security checks for vulnerabilities and agentic risk
This skill is a straightforward guide for using the local pixr image CLI, with disclosed persistent configuration behavior but no evidence of deception, exfiltration, or destructive actions.
Install this if you want Codex to operate the local pixr CLI. Be aware that commands such as setting the model, save directory, or profiles can persist under ~/.pixr and affect future pixr runs, and that generation uses your configured pixr/Gemini credentials and quota.
Skill selects an external model or provider that may use a different account or billing plan than the operator expects. Undisclosed model switches can cause unexpected cost or quota consumption.
pixr profile list pixr profile show social pixr profile init social pixr profile init social --model gemini-3.1-flash-image-preview --save-dir "~/Pictures/pixr/social"
Notes:
The skill encourages use of persistent home-directory state such as ~/.pixr/assets, ~/.pixr/config.json, and profile initialization, which can influence future runs outside the current session. In context this is expected CLI functionality, but without guardrails it enables an agent to create or change durable settings and assets that persist across tasks, making the behavior more dangerous than a purely ephemeral workflow.
- "generate an image with pixr"
- "edit an image with pixr"
- "create variations with pixr"
- "set the pixr model"
- "save images to a default folder"
- "use the refs from ~/.pixr/assets"
The skill enables implicit invocation but does not define a narrow, explicit trigger boundary. That means the agent may auto-select this skill for broad image-related requests and execute local CLI operations or configuration changes without sufficiently clear user intent, increasing the risk of unintended file writes, edits, or configuration manipulation.
The skill explicitly recommends commands that persistently modify user configuration under ~/.pixr, including saved model, save directory, and profile defaults, but does not instruct the agent to obtain user confirmation before making those changes. This creates a session-persistence risk because an agent may silently alter future CLI behavior and filesystem destinations beyond the current task.