name	copilot-cli-runtime
description	Internal helper contract for calling the copilot-companion runtime from Claude Code
user-invocable	false

Copilot Runtime

Use this skill only inside the copilot:copilot-rescue subagent.

Primary helper:

node "${CLAUDE_PLUGIN_ROOT}/scripts/copilot-companion.mjs" task "<raw arguments>"

Execution rules:

The rescue subagent is a forwarder, not an orchestrator. Its only job is to invoke task once and return that stdout unchanged.
Prefer the helper over hand-rolled git, direct Copilot CLI strings, or any other Bash activity.
Do not call setup, review, adversarial-review, status, result, or cancel from copilot:copilot-rescue.
Use task for every rescue request, including diagnosis, planning, research, and explicit fix requests.
You may use the copilot-prompting skill to rewrite the user's request into a tighter Copilot prompt before the single task call.
That prompt drafting is the only Claude-side work allowed. Do not inspect the repo, solve the task yourself, or add independent analysis outside the forwarded prompt text.
Leave --effort unset unless the user explicitly requests a specific effort.
Leave model unset by default. Add --model only when the user explicitly asks for one.
Default to a write-capable Copilot run by adding --write unless the user explicitly asks for read-only behavior or only wants review, diagnosis, or research without edits.

Command selection:

Use exactly one task invocation per rescue handoff.
If the forwarded request includes --background or --wait, treat that as Claude-side execution control only. Strip it before calling task, and do not treat it as part of the natural-language task text.
If the forwarded request includes --model, pass it through to task unchanged. Accepted aliases: fast → claude-opus-4.6-fast, opus → claude-opus-4.7, sonnet → claude-sonnet-4.6, haiku → claude-haiku-4.5, gpt → gpt-5.5, codex → gpt-5.3-codex, auto → Copilot's auto-model selection (GA 2026-04-17). Any concrete Copilot model name works too; it passes through unchanged.
If the forwarded request includes --effort, pass it through to task. Copilot CLI 1.0.11+ has a native --effort=<low|medium|high|xhigh> flag — the companion forwards it verbatim. The plugin's none/minimal aliases collapse to low at spawn time for codex-plugin-cc command parity.
--model and --effort are independent. Pass either or both; Copilot's runtime applies them without conflict. (Pre-v0.0.16 the plugin internally mapped effort to a Claude model tier and emitted a "ignored because --model was passed" stderr notice when both were set — that mapping was removed in v0.0.16.)
If the forwarded request includes --resume, strip that token from the task text and add --resume-last.
If the forwarded request includes --fresh, strip that token from the task text and do not add --resume-last.
--resume: always use task --resume-last, even if the request text is ambiguous.
--fresh: always use a fresh task run, even if the request sounds like a follow-up.
--effort: accepted values are none, minimal, low, medium, high, xhigh.
task --resume-last: internal helper for "keep going", "resume", "apply the top fix", or "dig deeper" after a previous rescue run. Within a single Claude session the companion reuses the cached Copilot session ID over ACP (full streaming progress). Across Claude sessions there's no in-broker fallback — render.mjs emits a copilot --continue ${sessionId} hint string so users can resume via the Copilot CLI directly (the in-plugin Path B subprocess fallback was descoped in v1).

Safety rules:

Default to write-capable Copilot work in copilot:copilot-rescue unless the user explicitly asks for read-only behavior.
Preserve the user's task text as-is apart from stripping routing flags.
Do not inspect the repository, read files, grep, monitor progress, poll status, fetch results, cancel jobs, summarize output, or do any follow-up work of your own.
Return the stdout of the task command exactly as-is.
If the Bash call fails or Copilot cannot be invoked, return nothing.

Copilot Runtime

Use this skill only inside the copilot:copilot-rescue subagent.

Primary helper:

node "${CLAUDE_PLUGIN_ROOT}/scripts/copilot-companion.mjs" task "<raw arguments>"

Execution rules:

The rescue subagent is a forwarder, not an orchestrator. Its only job is to invoke task once and return that stdout unchanged.

Prefer the helper over hand-rolled git, direct Copilot CLI strings, or any other Bash activity.

Do not call setup, review, adversarial-review, status, result, or cancel from copilot:copilot-rescue.

Use task for every rescue request, including diagnosis, planning, research, and explicit fix requests.

You may use the copilot-prompting skill to rewrite the user's request into a tighter Copilot prompt before the single task call.

That prompt drafting is the only Claude-side work allowed. Do not inspect the repo, solve the task yourself, or add independent analysis outside the forwarded prompt text.

Leave --effort unset unless the user explicitly requests a specific effort.

Leave model unset by default. Add --model only when the user explicitly asks for one.

Default to a write-capable Copilot run by adding --write unless the user explicitly asks for read-only behavior or only wants review, diagnosis, or research without edits.

Command selection:

Use exactly one task invocation per rescue handoff.

If the forwarded request includes --background or --wait, treat that as Claude-side execution control only. Strip it before calling task, and do not treat it as part of the natural-language task text.

If the forwarded request includes --model, pass it through to task unchanged. Accepted aliases: fast → claude-opus-4.6-fast, opus → claude-opus-4.7, sonnet → claude-sonnet-4.6, haiku → claude-haiku-4.5, gpt → gpt-5.5, codex → gpt-5.3-codex, auto → Copilot's auto-model selection (GA 2026-04-17). Any concrete Copilot model name works too; it passes through unchanged.

If the forwarded request includes --effort, pass it through to task. Copilot CLI 1.0.11+ has a native --effort=<low|medium|high|xhigh> flag — the companion forwards it verbatim. The plugin's none/minimal aliases collapse to low at spawn time for codex-plugin-cc command parity.

--model and --effort are independent. Pass either or both; Copilot's runtime applies them without conflict. (Pre-v0.0.16 the plugin internally mapped effort to a Claude model tier and emitted a "ignored because --model was passed" stderr notice when both were set — that mapping was removed in v0.0.16.)

If the forwarded request includes --resume, strip that token from the task text and add --resume-last.

If the forwarded request includes --fresh, strip that token from the task text and do not add --resume-last.

--resume: always use task --resume-last, even if the request text is ambiguous.

--fresh: always use a fresh task run, even if the request sounds like a follow-up.

--effort: accepted values are none, minimal, low, medium, high, xhigh.

task --resume-last: internal helper for "keep going", "resume", "apply the top fix", or "dig deeper" after a previous rescue run. Within a single Claude session the companion reuses the cached Copilot session ID over ACP (full streaming progress). Across Claude sessions there's no in-broker fallback — render.mjs emits a copilot --continue ${sessionId} hint string so users can resume via the Copilot CLI directly (the in-plugin Path B subprocess fallback was descoped in v1).

Safety rules:

Default to write-capable Copilot work in copilot:copilot-rescue unless the user explicitly asks for read-only behavior.

Preserve the user's task text as-is apart from stripping routing flags.

Do not inspect the repository, read files, grep, monitor progress, poll status, fetch results, cancel jobs, summarize output, or do any follow-up work of your own.

Return the stdout of the task command exactly as-is.

If the Bash call fails or Copilot cannot be invoked, return nothing.

copilot-cli-runtime

Copilot Runtime

Mehr aus diesem Repository

Mehr aus diesem Repository

Copilot Runtime