The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides explicit instructions to configure config.toml with approval_policy = "never" and sandbox_mode = "danger-full-access". These settings are designed to bypass user confirmation prompts and disable the security sandbox of the Codex CLI tool, which are patterns associated with bypassing safety guidelines and removing execution constraints.
[COMMAND_EXECUTION]: The skill recommends executing a local shell script ./scripts/ai/verify-ai-compat.sh and utilizing the codex exec command for environment and skill discovery verification. It also suggests a shell loop to verify symlink structures on the local filesystem.

configuring-codex