using-superpowers

To force a skill check before an agent replies, you need mandatory skill invocation orchestration. This treats skill applicability as a non-zero probability, triggering the skill tool before any response or clarifying question is sent.

Agent instructions in CLAUDE.md or similar files override default system behavior through priority resolution. User-provided directives stay highest priority, while superpower skills override default system behavior where conflicts occur.

Yes, agent orchestration works with Copilot CLI and Gemini CLI. It provides multi-platform activation guidance, explaining how to trigger skills across various environments using the appropriate platform-specific tool or activation mechanism.

Limitations of mandatory skill invocation include execution guardrails where it skips itself when dispatched as a subagent. It also instructs against reading skill files via the Read tool, which is a specific constraint for Claude Code environments.

You need workflow guardrails for agent orchestration during interactive coding and planning conversations. They ensure the correct workflow is followed by detecting applicable skills and enforcing mandatory pre-response checks before any user-facing action.