VeryGoodOpenSource / VeryGoodOpenSource/vgv-wingspan

feat: add support for other AI platforms other than Claude Code

Open
#66 0 comments 5 reactions 1 assignee View on GitHub

@ryzizub is already working on this.

Since Jul 21, 2026.

feature p1
Dominant language
Shell
Stars
105
Forks
15
Avg merge
13h 12m
Merged PRs (30d)
2

Description

Description

Epic for the open-standard / multi-harness effort.

Strategy: publish skills to the open agent-skills standard (.agents/skills/ convergence + skills.sh registry, https://agentskills.io/home) instead of maintaining per-harness plugin manifests. Tiers:

  • Tier 1 — Claude Code: full plugin (skills + agents + recommend-plugins hook + MCP). Primary, always supported.
  • Tier 2 (wave 1) — Codex + Gemini CLI + OpenCode: skills verified + CI-tested. All three read .agents/skills/ directly, support MCP (context7), and support subagents. Codex and Gemini ship Claude-compatible hook systems (JSON-over-stdin), so the recommend-plugins hook ports with adaptation; OpenCode uses a JS/TS plugin API, so its hook is a rewrite.
  • Experimental: every other harness that reads .agents/skills/ (Cursor, Antigravity, Copilot, ...) — skills work for free, no dedicated CI or guarantees.
Subagents port to tier 2

Wingspan's power is its subagents (analysis / codebase-review / quality-review / research agents, and the review/plan panels). Good news: both tier-2 harnesses now ship full subagent support with isolated context, curated tools, MCP, and parallel dispatch — so these flows port rather than degrade.

  • Gemini CLI (subagents since v0.36): custom agents are Markdown + YAML frontmatter in .gemini/agents — nearly identical to our existing agents/*.md, so close to a copy.
  • Codex (subagents GA March 2026): custom agents are TOML in ~/.codex/agents/, driven by spawn_agent / wait_agent primitives and /agent. Requires an md -> TOML conversion of our agent definitions.
  • OpenCode: markdown subagents (mode: subagent, invoked via @mention) — close to our existing agents/*.md.

Remaining work is format conversion, not reimplementation of the workflow.

Linked under this epic
  • #224 release in skills.sh
  • #225 support AGENTS standard
  • #226 Codex support (tier 2)
  • #227 Gemini CLI support (tier 2)
  • #229 OpenCode support (tier 2)
  • #228 Phase 0 portability prep — flatten namespacing + cross-plugin collision audit ($ARGUMENTS / ${CLAUDE_SKILL_DIR} fallbacks partly landed via #220), context7 MCP guard, reserved-command audit, agent-definition format conversion (md -> TOML for Codex; md+frontmatter for Gemini; md agents for OpenCode) (blocks the rest)
Related
  • PR #220 (GitHub Copilot CLI dual-manifest) — pending decision: merge-and-freeze as the last per-harness manifest, or close in favor of open-standard-only. Its $ARGUMENTS / ${CLAUDE_SKILL_DIR} fallbacks and host-aware hook are reusable regardless.
  • Closed prior attempts: #16, #22 (Cursor).
Requirements

No response

Additional Context

No response

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.