Add Adoption of Spec Driven Development Using Superpowers Plugin

AI helpers such as opencode make it easy to change the codebase, but without a shared workflow every agent session invents its own process: some write tests, most do not, and plans are written once and then lost. The result is uneven quality, requests that drift from the original intent, and work that is hard to review. The specs produced by the workflow would also be lost: they are useful documentation, so they must be converted to asciidoc and published in the Antora docs (following .agents/rules/specs.md) instead of being thrown away. How should we drive the work of AI helpers so that it is spec-driven, testable, repeatable, and its specs end up in the published docs?

Chosen option: "Adopt superpowers", because it ships a complete, battle-tested workflow as reusable skills instead of prose, and it registers and updates itself through opencode’s plugin manager. Positive consequences: bundled skills (brainstorming, plan writing, TDD, verification) keep agent work spec-driven and consistent across sessions, new conventions are discoverable as skills rather than buried in docs, and the written specs become long-lived documentation once converted to antora pages under antora/modules/specs/ per .agents/rules/specs.md. Negative consequences: agents lean on guidance code that lives in a third-party repository, and the workflow adds overhead to trivially small tasks.

Adopting superpowers is the best option, because it gives a maintained, coherent process for free and updates through the opencode plugin ecosystem. Hand-writing the rules in AGENTS.md was not the best, because rules as prose are easy to ignore and drift out of date. Sticking with ad hoc prompts would keep the quality problems that motivated this ADR unsolved.

See https://github.com/obra/superpowers/blob/main/docs/README.opencode.md for the installation and usage instructions for opencode.