The Agent Skills Directory

[COMMAND_EXECUTION]: The validation section in SKILL.md contains commands to execute local Python linting and validation scripts (e.g., python3 $CODEX_HOME/skills/senior-skill-architect/scripts/lint_production_skill.py). These are standard development tools used for internal skill verification.\n- [SAFE]: The skill instructions provide robust safety constraints, requiring the agent to prioritize planning and dry-runs. It explicitly forbids the execution of irreversible actions (like purchasing domains or provisioning servers) and the printing of credentials without recorded human approval.\n- [SAFE]: (Indirect Prompt Injection) The skill coordinates several infrastructure-related tools. It processes external project requirements (Ingestion point) but mitigates risk through explicit 'dry-run' and 'approval' boundary markers. Its capability inventory includes orchestrating deployment and provisioning tools, which is secured by mandatory human-in-the-loop sanitization/approval gates before any command execution occurs.

infra-launch-orchestrator