Security audit
Security checks for vulnerabilities and agentic risk
This skill openly makes Codex more autonomous for coding tasks, but it does not contain hidden code, persistence, exfiltration, or unrelated privileges.
Install this only if you want Codex to proceed through normal coding steps with fewer checkpoints. Review the trigger phrases and avoid enabling it for work where every file edit, dependency install, long-running command, or local server launch should require a separate confirmation.
The activation description uses broad, natural-language trigger phrases such as 'nao pare' and 'rode ate o final' that can easily appear in ordinary user requests. This increases the chance the skill activates unintentionally and shifts the agent into a more autonomous mode without explicit, informed user opt-in.
This instruction explicitly authorizes the agent to make routine choices, run commands, edit files, and verify results 'without asking for avoidable confirmations.' In a skill designed for autonomous execution, that materially reduces human oversight and can lead to unintended code changes, command execution, or persistence in risky workflows if the original request is ambiguous or manipulated.
## Operating Mode
Execute the requested task end to end. Treat the user's latest prompt as permission to make routine choices, run commands, edit files, wait for output, and verify results without asking for avoidable confirmations.
## Rules
The confirmation policy normalizes proceeding without asking across file reads, dependency installation, test/build execution, file edits, and waiting/polling. While convenient, this broad delegation expands the blast radius of accidental activation and makes the skill context more dangerous because it authorizes real system changes and code execution under implied consent.
## Confirmation Policy
Proceed without asking when the action is normal for the task:
- reading files, searching the repo, inspecting logs, checking status
- installing local dependencies already required by the project
The default prompt instructs the agent to run tasks end to end without avoidable confirmations, which broadens activation toward autonomous execution without defining concrete safety boundaries or narrow trigger conditions. In the context of a run-to-completion skill, ambiguous activation increases the chance the agent will suppress useful checkpoints before risky actions such as executing commands, modifying files, or waiting on long-running operations.
The description hard-codes Portuguese trigger phrases without documenting language scope, user consent, or why multilingual triggers are needed. This can surprise users and increase accidental activation risk, especially in mixed-language environments where the user may not realize these phrases enable an autonomous workflow.