[prompt-analysis] Copilot PR Prompt Analysis - 2026-07-04

[github-actions[bot]]

Summary

Analysis Period: Last 30 days
Total PRs: 1000 | Merged: 798 (81.1%) | Closed: 186 (18.9%) | Open: 16

Prompt Categories and Success Rates

Category	Total	Merged	Closed	Success Rate
Bug Fix	801	642	145	81.6%
Feature Addition	167	129	36	78.2%
Refactoring	9	8	1	88.9%
Documentation	7	7	0	100.0%
Testing	7	6	1	85.7%
Other	9	6	3	66.7%

Prompt Analysis

✅ Successful Prompt Patterns

Common characteristics in merged PRs:

• Average prompt length: 192 words
• Most common keywords: workflow, actions, agent, runs, output, safe, test, json
• Action verbs used: fix (222), add (141), refactor (35), update (31), test (24)

Top verb success note: fix appears in 222 merged titles vs 44 closed — a 5:1 ratio, indicating precise bug-targeting prompts strongly outperform vague prompts.

Example successful prompts:

1. PR docs(spec): fix parser and workflow README accuracy issues #43394 — docs(spec): fix parser and workflow README accuracy issues

   "Corrects four spec-implementation mismatches flagged by the 2026-07-04 Package Specification Audit..."
   → Merged ✅

2. PR require-await-core-summary-write: tighten isCoreLikeIdentifier to exact alias set #43382 — require-await-core-summary-write: tighten isCoreLikeIdentifier to exact alias set

   "isCoreLikeIdentifier used /^core/i, which over-matches any identifier starting with core..."
   → Merged ✅

3. PR deps: bump bubbletea v2.0.8, lipgloss v2.0.5 #43375 — deps: bump bubbletea v2.0.8, lipgloss v2.0.5

   "Patch bumps for two Charm rendering dependencies — both updates are emoji rendering bug fixes..."
   → Merged ✅

❌ Unsuccessful Prompt Patterns

Common characteristics in closed PRs:

• Average prompt length: 190 words (nearly identical to merged — length alone is not predictive)
• Frequent signals: WIP in title, broad scope words (introduce, scaffold, detect, feat:), vague goals
• Issue keywords elevated in closed: model, issue, shared, docs, outputs, claude

Example unsuccessful prompts:

1. PR Add daily HTML-change → Playwright test generator workflow #43421 — Add daily HTML-change → Playwright test generator workflow

   "Introduces html-web-test-generator, a scheduled agentic workflow that detects HTML file changes..."
   → Closed ❌

2. PR feat: AWEngine runtime adapter, aw_harness.cjs scaffold, and threat-detection-suppress frontmatter #43416 — feat: AWEngine runtime adapter, aw_harness.cjs scaffold, and threat-detection-suppress frontmatter

   "Implements three P0/P1 items from the 2026-07-04 SPDD spec plan..."
   → Closed ❌

3. PR [WIP] Add gh-aw-detection: true to audit workflow scaffold #43402 — [WIP] Add gh-aw-detection: true to audit workflow scaffold

   "Thanks for asking me to work on this. I will get started on it..."
   → Closed ❌

Key Insights

• Precision beats breadth: Bug-fix PRs (81.6% success) and refactoring PRs (88.9%) outperform broad feature PRs (78.2%). Prompts that name the specific code path, function, or symptom fare much better than ones describing a new system to build.
• Documentation PRs are safest: 100% merge rate (7/7). Prompts asking for doc corrections, spec alignment, or README fixes virtually never get rejected.
• WIP signals and multi-scope PRs are risky: Closed PRs often bundle multiple unrelated changes (AWEngine adapter + aw_harness + frontmatter) or declare [WIP] intent. Single-concern prompts merge at a significantly higher rate.
• Prompt length is not a differentiator: Merged (192 words avg) and closed (190 words avg) prompts are nearly identical in length — what matters is specificity and scope, not verbosity.

Recommendations

Based on today's analysis:

1. DO — Reference the specific file, function, or symptom: fix parser over-match in isCoreLikeIdentifier beats improve identifier matching.
2. DO — Keep each PR focused on a single concern; avoid bundling feat A + feat B + feat C in one prompt.
3. DO — Use conventional commit prefixes (fix:, docs:, deps:) — they correlate with higher merge rates by signaling bounded scope.
4. AVOID — Opening PRs with [WIP] in the title; these are rarely merged without human rework.
5. AVOID — Feature prompts that introduce a wholly new workflow or system without prior design alignment; these close at 21.8% vs docs at 0%.

Historical Trends

Date	Total PRs	Merged	Success Rate	Top Category
2026-07-04	1000	798	81.1%	Bug Fix
2026-07-03	1000	801	81.8%	Bug Fix

Trend: Success rate is stable (~81%), with bug-fix dominating the workload. Feature PRs shrank from 282 → 167 day-over-day, while bug-fix PRs grew from 650 → 801, suggesting more targeted fix-oriented prompting is occurring.

---

References:

• §28719391195

Generated by 🔍 Copilot PR Prompt Pattern Analysis · 43.8 AIC · ⌖ 25.2 AIC · ⊞ 7.8K · ◷

• expires on Jul 5, 2026, 1:04 PM UTC-08:00