القائمة
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.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
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 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.
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.
استنادا إلى تصنيف SOC المهني
| name | test-copilot-runtime |
| description | 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. |
How to spawn the copilot runtime against an isolated test server. Assumes a server from /launch-grackle (with GRACKLE_URL + GRACKLE_API_KEY exported).
| Model | Works? | Notes |
|---|---|---|
claude-sonnet-4.5 | ✅ confirmed live | The @github/copilot-sdk README default; spawns and runs |
gpt-5 | ⚠️ likely | Documented for Copilot CLI; not individually confirmed here |
claude-sonnet-4 | ⚠️ likely | From GitHub's supported-models docs |
claude-haiku-4.5 | ⚠️ likely | From GitHub's supported-models docs |
gpt-4o | ❌ does NOT work | Model "gpt-4o" is not available. GPT-4o is intentionally absent from the Copilot CLI/SDK (only in Copilot Chat + the VS Code extension) |
⚠️
grackle runtimesadvertisesgpt-4ofor copilot — that is wrong/unavailable for the CLI. Useclaude-sonnet-4.5. (Catalog entry:packages/common/src/runtime-catalog.ts; worth fixing — see the runtime-catalog.)
grackle persona create "Copilot Tester" --runtime copilot --model claude-sonnet-4.5 \
--prompt "You are a test agent. When asked to run a command, use your shell tool to run exactly that one command and nothing else."
grackle spawn local "Use your shell tool to run exactly this one command, then stop: cat /nonexistent_file_xyz" --persona copilot-tester
Copilot surfaces real per-tool outcomes on the tool.execution_complete event (data.success / data.error), which the adapter lifts to the first-class toolError field (packages/runtime-copilot/src/copilot.ts).
Confirmed live (#1362): a failing tool execution persisted as content: {"is_ok":false,...} and "tool_error":true. Copilot is the cleanest way to exercise the real tool-failure path end-to-end (native claude-code can't — see /test-claude-runtime).
$GRACKLE_HOME/.grackle/logs/<session-id>/stream.jsonl
tool_result content = {is_ok, content, past_tense_message}; failures carry "tool_error":true.
/launch-grackle, /test-claude-runtime, /test-codex-runtime, /test-acp-runtime