| name | copilot-cli |
| description | Delegates large multi-step work to GitHub Copilot CLI like a subagent—include full goal, prior decisions, scope, and constraints in prompts so isolated sessions stay aligned with the main thread (context handoff per Cognition-style delegation). Use for heavy edits, exploration, codebase research, @explore, @general-purpose, or @code-review. Skip for trivial one-shot tasks or when everything is already in context. |
| allowed-tools | ["Bash","Read","Write","Grep","Glob"] |
GitHub Copilot CLI (subagent/task delegation)
Use GitHub Copilot CLI to run a separate long-horizon pass over the repo: multi-step implementation, broad refactors, batch file writes, or deep exploration—similar to handing a task to a subagent. You stay orchestrator: smaller prompts, less context burn, often lower spend than doing the same work entirely in-session.
Inside Copilot CLI, agents are specialists. Force one using the --agent flag in non-interactive prompts. Core specialists include explore, general-purpose, task, research, and code-review.
When to use Copilot CLI
- Large or multi-step work: several files, phases, or checkpoints (feature slice, migration, test suite, docs sweep).
- Heavy code generation or editing: Copilot drives
--yolo tool use while you summarize outcomes and merge.
- Parallel mental lane: you continue planning or reviewing while Copilot runs a bounded task (optionally background—see reference.md).
- Built-in specialists:
explore for architecture / dependency mapping; general-purpose for heavy multi-file subtasks with isolated context; code-review for review-style analysis.
- User explicitly asks for Copilot or “use Copilot for this.”
When not to use
- Small / single-step tasks answerable with one or two edits or a short explanation.
- Tight feedback loops where the user wants rapid back-and-forth refinement in one thread.
- Secrets or policy-sensitive flows—avoid piping credentials; redact before delegating.
- Already-loaded context where duplicating the whole plan adds no value—handle locally.
- Low ROI (Return on Investment): If the task is "needle-in-a-haystack" (requires high precision over a single line) or if the time to compose the Handoff Table exceeds the time to simply edit the file locally. Delegation should only be used when the "mental offloading" outweighs the "handoff overhead."
Delegation and context (critical)
Isolated subagent context saves tokens but splits the story: Copilot does not see your full thread. Poor handoffs cause misread subtasks, conflicting assumptions (stack, style, APIs), and wasted edits—see Cognition’s discussion of sharing context and implicit decisions in multi-step setups.
When composing the single Copilot prompt (using -p), treat it as passing enough shared state, not just a title:
| Include | Why |
|---|
| Original goal | Same north star as the user—not only the immediate micro-task. |
| Decisions already made | Framework, patterns, naming, auth approach, “use X not Y”—anything that would otherwise be guessed wrong. |
| Scope | Paths, modules, env (@ FILENAME), and explicit out of scope / do-not-touch areas. |
| Constraints | Performance, a11y, compatibility, review gates, “no new deps,” etc. |
| Verification | Explicit command (e.g. npm test, lint) the subagent must run and pass before returning. |
| Expected output | e.g. “summarize then list files changed,” “report only—no edits,” or “apply edits with minimal diff.” |
After Copilot returns, pull decisions and constraints back into the main thread (what it assumed, what it changed, open risks)—so the next step or a follow-up copilot call does not contradict earlier work. Prefer sequential delegations with explicit carry-over over parallel runs that might diverge unless they share the same briefing.
If the delegation would need a long transcript to be safe, summarize the relevant parts into the prompt (compressed “state of the union”) rather than a one-line subtask.
Model Selection & Discovery (Mandatory)
MANDATORY: Search the web for latest Copilot model names/aliases and pricing before selecting a model. You should also consult Artificial Analysis for the most up-to-date benchmarks, pricing, and model performance data. Availability of specific high-end models varies by account and region.
- Default (Simple Tasks):
gpt-5.4-mini (Inexpensive, fast, ideal for simple edits and exploration).
- Heavy Tasks:
gpt-5.4 (Current high-performance model for complex reasoning and refactoring).
- Strategy: Always default to the "mini" variant to save credits. Only escalate to
gpt-5.4 for critical tasks after verifying latest version and cost via search.
Programmatic usage (required)
You MUST use Copilot CLI programmatically. Do NOT start interactive sessions.
| Requirement | Flag |
|---|
| Non-interactive | -p "prompt" or --prompt "prompt" |
| Silent output | -s or --silent (removes stats/decorations) |
| Auto-approval | --yolo (or --allow-all-tools to avoid hangs) |
| Model Selection | --model [model_name] |
| Reasoning Effort | `--effort [low |
Command pattern
copilot -p "[prompt with @ FILENAME as needed]" --yolo -s --model gpt-5.4-mini 2>&1
With specific agent:
copilot -p "[task]" --agent general-purpose --yolo -s 2>&1
After Copilot returns
- Review diffs and security-sensitive areas (XSS, injection, auth)—do not merge blindly.
- Run project checks (
lint, test, typecheck) as appropriate.
- Compress what you send upstream: summarize results for the user instead of pasting huge logs unless asked.
- Reconcile context: note decisions, files touched, and remaining risks so the main session stays aligned.
Quick prompts
- Delegate implementation:
copilot -p "GOAL: [goal] | DECISIONS: [decisions] | SCOPE: [paths] | CONSTRAINTS: [constraints] | VERIFICATION: [test_command] | OUTPUT: [format]" --yolo -s
- Investigate:
copilot -p "GOAL: Map how [feature] works | SCOPE: [paths] | OUTPUT: concise file:line map" --agent explore --yolo -s
- Audit:
copilot -p "GOAL: Audit the codebase for security issues | SCOPE: [paths] | OUTPUT: security report" --agent code-review --yolo -s
More detail
