[prompt-analysis] Copilot PR Prompt Analysis - 2026-07-08

[github-actions[bot]]

Summary

Analysis Period: Last 30 days (up to 2026-07-08)
Total PRs: 1,000 | Merged: 795 (81.0%) | Closed (not merged): 195 (19.7%) | Open: 10 (1.0%)

---

Prompt Categories and Success Rates

Category	Total	Merged	Closed	Success Rate
Bug Fix	794	639	148	81.2%
Documentation	48	39	9	81.2%
Refactoring	30	24	6	80.0%
Testing	60	43	14	75.4%
Feature Addition	51	38	13	74.5%
Other	17	12	5	70.6%

Bug Fix and Documentation prompts are the top performers at 81.2% success; Feature and Testing prompts lag slightly behind.

---

✅ Successful Prompt Patterns

Common characteristics in merged PRs:

• Average prompt length: 181.9 words
• Most common keywords: workflow, actions, github, agent, engine, copilot
• 97.1% use bullet/checklist format (vs 91.8% in closed)
• 82.1% include code blocks (vs 72.8% in closed)
• Only 1.4% have [WIP] in the title (vs 12.3% in closed)
• 88.9% reference specific files/paths

Example Successful Prompts

PR #44308 – docs(parser): document missing internal dependencies in README

`pkg/parser/README.md` was missing three internal dependencies from its `## Dependencies` section, flagged by the spec-linter...
→ Merged ✅

PR #44288 – add smoke-copilot-small workflow to validate model: small alias resolution

The Copilot CLI validates model names before making API calls, rejecting meta-aliases like 'small' as "retired or unsupported"...
→ Merged ✅

PR #44286 – fix: default sink-visibility to "public" when repository visibility check fails

When the 'determine-automatic-lockdown' step fails to fetch repository visibility, 'sink-visibility' was set to "unknown"...
→ Merged ✅

PR #44282 – Compiler: auto-grant pull-requests:read to pre_activation for decentralized label_command

When 'label_command' uses 'strategy: decentralized' with 'pull_request' events, the compiled 'pre_activation' job inherits...
→ Merged ✅

---

❌ Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• Average prompt length: 182.4 words (nearly identical to merged — length alone doesn't predict outcome)
• Most common keywords: model, issue, chore, bump, deps, routine
• 12.3% have [WIP] in title (8.8× higher than merged PRs)
• Only 72.8% use code blocks (vs 82.1% for merged)
• Dependency-bump PRs and routine chores with vague justification appear frequently

Example Unsuccessful Prompts

PR #44269 – docs: add golang.org/x/crypto Go Fan module summary

Go Fan reviewed 'golang.org/x/crypto' and confirmed the dependency is healthy — pinned at the latest tag (v0.53.0)...
→ Closed ❌

PR #44267 – chore(deps): bump charmbracelet/x/exp/golden pseudo-version

Pseudo-version bump in the charmbracelet/x monorepo. The 3 commits between versions make zero code changes...
→ Closed ❌

PR #44239 – chore: bump Claude Code, Copilot CLI, Codex

Routine patch/minor version bumps for three agentic CLI tools...
→ Closed ❌

---

Key Insights

• WIP title is a strong failure signal: PRs titled [WIP] are closed at 8.8× the rate of merged PRs (12.3% vs 1.4%). Avoid submitting prompts that produce WIP PRs unless specifically intended.
• Code blocks correlate with success: Merged PRs include code blocks at an 82.1% rate vs 72.8% for closed PRs — prompts that reference concrete code snippets or examples get merged more reliably.
• Checklist/bullet format matters: 97.1% of merged PR bodies use structured bullet or checklist format, suggesting that prompts decomposed into discrete steps produce cleaner, more reviewable PRs.
• Routine chores underperform: chore(deps) and routine version-bump PRs dominate the closed category — consider reviewing whether these should be auto-approved or skipped entirely.

---

Recommendations

1. DO: Frame prompts around a concrete problem with specific file references and expected behavior — e.g., "When X happens in pkg/foo/bar.go, Y should occur instead of Z"
2. DO: Decompose work into checklist steps in the PR body — structured task lists are present in 97% of merged PRs
3. DO: Include a code block showing the before/after or the relevant snippet — code-block PRs merge at 82% vs 73%
4. AVOID: Submitting PRs marked [WIP] — they are 8.8× more likely to be closed
5. AVOID: Vague routine chore descriptions like "routine version bumps" — provide specific rationale for why the change is needed now

---

Historical Trends (Last 6 Days)

Date	Total PRs	Merged	Success Rate	Top Category
2026-07-08	1,000	795	81.0%	Bug Fix (81.2%)
2026-07-07	1,000	792	79.2%	Bug Fix (80.8%)
2026-07-06	1,000	793	79.3%	Testing (83.4%)
2026-07-05	1,000	805	80.5%	Bug Fix (81.6%)
2026-07-04	1,000	798	79.8%	Refactor (88.9%)
2026-07-03	1,000	801	80.1%	Testing (88.2%)

Trend: Overall merge rate is stable in the 79–81% range. Today's 81.0% is the highest of the tracked period, matching a slight uptick in documentation-quality prompts and fewer WIP-titled submissions.

---

References:

• Workflow run: §28975162639

Generated by 🔍 Copilot PR Prompt Pattern Analysis · 49.6 AIC · ⌖ 19.7 AIC · ⊞ 7.4K · ◷

• expires on Jul 9, 2026, 1:06 PM UTC-08:00

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-07-09T21:06:49.841Z.

Closed by Workflow