菜单
Spawn the `codex` runtime against a test server. Use model `gpt-5.5` (requires Codex SDK >= 0.135.0). Run with /test-codex-runtime. Pairs with /launch-grackle.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Spawn the ACP-variant runtimes (claude-code-acp / codex-acp / copilot-acp). claude-code-acp validated; it surfaces real tool failures that native claude-code can't. Run with /test-acp-runtime. Pairs with /launch-grackle.
Spawn and exercise the native `claude-code` runtime against a test server, including the model names that work and the synthetic-tool-result gotcha. Run with /test-claude-runtime. Pairs with /launch-grackle.
Spawn and exercise the `copilot` runtime against a test server. CRITICAL: gpt-4o does NOT work — use claude-sonnet-4.5. Run with /test-copilot-runtime. Pairs with /launch-grackle.
Investigates code bugs and files GitHub issues.
Pure coordinator that decomposes work and delegates to subagents.
Verifies a PR is fully mergeable (CI green, no unresolved comments, approved) then merges it.
| name | test-codex-runtime |
| description | Spawn the `codex` runtime against a test server. Use model `gpt-5.5` (requires Codex SDK >= 0.135.0). Run with /test-codex-runtime. Pairs with /launch-grackle. |
How to spawn the codex runtime against an isolated test server. Assumes a server from /launch-grackle (with GRACKLE_URL + GRACKLE_API_KEY exported).
| Model | Works? | Notes |
|---|---|---|
gpt-5.5 | ✅ confirmed live | Needs Codex SDK >= 0.135.0 (the catalog pin). On the old 0.111.0 it errored "requires a newer version of Codex." |
o3, o3-mini, gpt-5-codex, gpt-5, gpt-5.1, gpt-5-mini | ❌ | "not supported when using Codex with a ChatGPT account" — these are account-gated (likely need OpenAI API-key auth or a higher tier), independent of SDK version |
The catalog (
packages/common/src/runtime-catalog.ts) advertisesgpt-5.5and pins@openai/codex-sdk ^0.135.0. Ifgrackle runtimesstill showso3, the catalog hasn't been rebuilt/installed.
History: the SDK was bumped ^0.111.0 → ^0.135.0 precisely to unlock gpt-5.5 (validated 2026-05-29). The runtime installer (runtime-installer.ts) auto-reinstalls when the catalog spec changes (it compares the persisted package spec), so the global ~/.grackle/runtimes/codex/ upgrades on the next codex spawn.
grackle persona create "Codex Tester" --runtime codex --model gpt-5.5 --prompt "You are a test agent."
grackle spawn local "Run exactly this one shell command and nothing else, then stop: cat /nonexistent_file_xyz" --persona codex-tester
The model string is passed verbatim to the SDK (packages/runtime-codex/src/codex.ts — threadOptions.model = this.model); Grackle does no validation, so any model the installed Codex accepts is valid.
Codex surfaces real per-tool outcomes by type — command_execution non-zero exit_code, file_change status, mcp_tool_call error — lifted to the first-class toolError field. Verified end-to-end: a failing command (cat /nonexistent → [exit 1]) persisted as content {is_ok:false} and "tool_error":true. A successful command omits tool_error.
$GRACKLE_HOME/.grackle/logs/<session-id>/stream.jsonl
tool_result content = {is_ok, content, past_tense_message}; failures carry "tool_error":true. Model-gating rejections show up as error + status: failed events (check the log to distinguish a rejection from a real run).
/launch-grackle, /test-claude-runtime, /test-copilot-runtime, /test-acp-runtime