Skip to content

Add CONTRIBUTING.md guidance on writing prompts for Copilot-assigned issues #53257

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
pelikhan merged 2 commits into from
Aug 17, 2026
Merged

Add CONTRIBUTING.md guidance on writing prompts for Copilot-assigned issues #53257

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f21b879
Initial plan
Copilot Aug 17, 2026
af2857c
Add prompt-writing guidance for Copilot-assigned issues to CONTRIBUTI…
Copilot Aug 17, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 13 additions & 0 deletions CONTRIBUTING.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -196,6 +196,19 @@ When filing issues with agentic plans:

**Quality of the agentic plan directly impacts implementation success.** Provide comprehensive, step-by-step instructions with specific details.

#### ✍️ Writing Prompts for Copilot-Assigned Issues

When an issue will be assigned to Copilot coding agent (or another SWE agent), how you write the task description measurably affects whether the resulting PR merges. Recurring "Copilot PR Prompt Pattern Analysis" reports (see [discussion #53203](https://github.com/github/gh-aw/discussions/53203)) show a consistent trend: merged PRs average **~151 words** per task prompt and name concrete files or subsystems (e.g. `pkg`, `engine`, a specific package), while closed/unsuccessful PRs average **~229 words** — roughly 50% longer — and skew toward open-ended, exploratory language without a concrete acceptance criterion.

To improve merge rate for agent-assigned work:

- **Be concise.** Aim for a short, focused prompt rather than a long, exploratory one. Prompts longer than ~200 words without a concrete acceptance criterion correlate with lower merge rates.
- **Name concrete files or subsystems.** Reference specific packages, files, or functions (e.g. `pkg/workflow`, `cmd/gh-aw`) instead of describing the problem in the abstract.
- **State an explicit acceptance criterion.** Describe what "done" looks like (e.g. "the CLI flag `--foo` is parsed and validated" or "add a test in `pkg/parser/parser_test.go` covering X") instead of open-ended language like "investigate and fix as appropriate."
- **Avoid pure exploration prompts.** Prompts that only ask the agent to "look into" or "figure out" a problem without a concrete target tend to result in longer, less-scoped PRs that are more likely to be closed.

Note: automated CVE/dependency-tracker PRs are excluded from this signal, since their close rate reflects triage policy rather than prompt quality.

### Code Quality Standards

Core team members and the agents they use follow these standards:
Expand Down
Loading