Pi agent runtime compatibility

[iamfj]

First of all, thank you for Spacedock — this project looks very useful, especially for structured multi-stage agent workflows.

I wanted to try it with the Pi agent runtime. After reading the repository, it seems Pi compatibility is not available yet (at least not currently wired/documented).

What I checked

From code/docs, runtime integration appears focused on Claude Code and Codex:

• README.md lists prerequisites as Claude Code or Codex CLI
• skills/first-officer/SKILL.md runtime adapter branches on:
  • CLAUDECODE (Claude)
  • CODEX_HOME (Codex)
• skills/ensign/SKILL.md follows the same runtime split
• Plugin metadata appears under .claude-plugin/* and .codex-plugin/*

Request

Would you be open to either:

1. Supporting Pi agent runtime in a future update, or
2. Adding a small runtime compatibility section in README (including current Pi status), so users can quickly understand what is supported today?

I’d also be happy to help support Pi integration if that would be useful.

[clkao]

Thanks for check this out on pi @iamfj! i did test birefly on pi a while back. since pi is intentionally minimum, there might be some heavy lifting getting the multi subagent support properly. contribution is welcome!

[clkao]

@iamfj i have pi+gpt-5.4 churned out something in 3 sessions. please check: #155

Since there are no native subagents, and prompt construction and spawning are a bit crude. we need to find a good way to hide them from the tui. also the e2e tests are currently pi-specific and we should probably merge them with other runtime.

let me know if that works for you!

PS: it was super fun working on this. I asked the (crippled) first officer in pi to pretend this already works with pi, by manually constructing subagent prompts and pretend all the constructs exist, and then implement the helpers

[clkao]

seems working well with pi-subagent