copilot-prompting

Copilot Prompting

Use this skill when copilot:copilot-rescue needs to send a task to GitHub Copilot.

Prompt Copilot like an operator, not a collaborator. Keep prompts compact and block-structured. State the task, the output contract, the follow-through defaults, and the small set of constraints that matter for this specific run.

Core Rules

• Prefer one clear task per Copilot run. Split unrelated asks into separate runs.
• Tell Copilot what done looks like. Do not assume it will infer the desired end state.
• Add explicit grounding and verification rules for any task where unsupported guesses would hurt quality.
• Prefer better prompt contracts over long natural-language explanations.
• Use XML tags consistently so the prompt has stable internal structure.

Default Prompt Recipe

• <task>: the concrete job and the relevant repository or failure context.
• <output_contract>: exact shape, ordering, and brevity requirements for the response.
• <follow_through_policy>: what Copilot should do by default instead of asking routine questions.
• <verification_loop>: required for debugging, implementation, or risky fixes.
• <grounding_rules>: required for review, research, or anything that could drift into unsupported claims.

When to Add Blocks

• Coding or debugging: add verification_loop, completeness_contract, and missing_context_gating.
• Review or adversarial review: add grounding_rules, output_contract, and a nudge to dig deeper on high-severity findings.
• Research or recommendation tasks: add research_mode and citation_rules.
• Write-capable tasks: add action_safety so Copilot stays narrow and avoids unrelated refactors.

How to Choose Prompt Shape

• Use the built-in review or adversarial-review commands when the job is reviewing local git changes. Those prompts already carry the review contract — do not re-specify them.
• Use task when the job is diagnosis, planning, research, or implementation and you need to control the prompt more directly.
• Use task --resume for follow-up instructions on the same Copilot session. Send only the delta instruction instead of restating the whole prompt unless the direction changed materially.

Working Rules

• Prefer explicit output contracts over vague nudges.
• The plugin always uses auto model — never suggest switching models. Tighten the prompt and verification rules if quality is insufficient.
• Ask Copilot for brief, outcome-based progress updates only when the task is long-running or tool-heavy.
• Keep claims anchored to observed evidence. If something is a hypothesis, say so explicitly.
• When writing tasks that will create files, name the exact target path in the prompt — Copilot respects explicit write targets more reliably than implicit location hints.

Prompt Assembly Checklist

1. Define the exact task and scope in <task>.
2. Choose the smallest output contract that still makes the answer easy to use.
3. Decide whether Copilot should keep going by default or stop for missing high-risk details.
4. Add verification, grounding, and safety tags only where the task needs them.
5. Remove redundant instructions before sending the prompt.

Rate Limit Awareness

GitHub Copilot has weekly premium request limits. If a task is likely to be long-running:

• Prefer --background so the session is not blocked.
• Avoid sending exploratory tasks that could be resolved with a quick grep or file read first.
• Do not retry automatically on rate-limit errors — surface the reset time to the user.
• The auto model runs against the free pool and is less likely to hit premium limits, but if it does, report it and stop.