Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when you need to review a plan, OpenSpec change, specification, or implementation proposal for breaking-change risk across commands, skills, generated outputs, XML sources, README/docs, tests, CI, APIs, schemas, configuration, data, migration, and release guidance. This should trigger for requests such as Review breaking changes in this spec; Check compatibility risks; Avoid breaking changes in this OpenSpec change; Review migration impact before release; Assess command and skill compatibili
.claude/skills/jabrena-056-design-avoid-breaking-changes/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✓→✗ | ▼ Worse | -60% | 0% |
| case-08 | ✓→✗ | ▼ Worse | -33% | 0% |
| case-11 | ✓→✗ | ▼ Worse | -50% | 0% |
| case-13 | ✓→✗ | ▼ Worse | -61% | 0% |
| case-15 | ✓→✗ | ▼ Worse | -54% | 0% |
Review planned repository changes for breaking-change and compatibility risk before implementation or release promotion. This is an interactive SKILL.
What is covered in this Skill?
.agents/skills, public skills/, .cursor/commands, .cursor/rules, and docs/Keep the review evidence-based, read-only, and scoped to compatibility risk unless the user explicitly asks for implementation.
references/056-design-avoid-breaking-changes.md before producing compatibility guidanceRead references/056-design-avoid-breaking-changes.md, then identify the source artifacts to review, their authority, and whether the request is read-only review or approved implementation.
Check commands, skills, generated outputs, XML/source ownership, README/docs, tests, CI/build gates, external contracts, runtime behavior, data/configuration contracts, and release or migration guidance relevant to the proposed change.
Classify each finding as BREAKING, POTENTIALLY BREAKING, NON-BREAKING, or UNKNOWN. Include the affected users, artifacts, contracts, and validation evidence or missing evidence.
For each confirmed or potential risk, recommend deprecation, compatibility windows, migration notes, release-note updates, generator checks, focused tests, or owner decisions.
Produce a concise compatibility report with reviewed sources, risk summary by surface, severity-ranked findings, recommended validation, and a clear no-risk statement when no breaking-change risks are found.
For detailed guidance, examples, and constraints, see references/056-design-avoid-breaking-changes.md.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | fail→fail | 7,247 | 18,594 | +157% | 1 | 1 | 0% | 1,293 | 3,696 | +186% | 0 | 0 | — |
case-05 | pass→pass | 14,650 | 11,746 | -20% | 1 | 1 | 0% | 2,477 | 2,753 | +11% | 0 | 0 | — |
case-01 | fail→fail | 17,074 | 5,685 | -67% | 1 | 1 | 0% | 2,572 | 1,071 | -58% | 0 | 0 | — |
case-02 | fail→fail | 20,496 | 5,367 | -74% | 1 | 1 | 0% | 2,856 | 1,028 | -64% | 0 | 0 | — |
case-03 | fail→fail | 18,074 | 4,578 | -75% | 1 | 1 | 0% | 2,701 | 1,120 | -59% | 0 | 0 | — |
case-06 | pass→fail | 13,540 | 5,654 | -58% | 1 | 1 | 0% | 2,593 | 1,029 | -60% | 0 | 0 | — |
case-07 | fail→fail | 7,143 | 6,769 | -5% | 1 | 1 | 0% | 667 | 1,139 | +71% | 0 | 0 | — |
case-08 | pass→fail | 14,469 | 6,349 | -56% | 1 | 1 | 0% | 1,626 | 1,091 | -33% | 0 | 0 | — |
case-09 | fail→fail | 14,458 | 6,803 | -53% | 1 | 1 | 0% | 2,185 | 1,030 | -53% | 0 | 0 | — |
case-10 | fail→fail | 13,268 | 17,354 | +31% | 1 | 1 | 0% | 2,002 | 3,490 | +74% | 0 | 0 | — |
case-11 | pass→fail | 26,998 | 4,675 | -83% | 1 | 1 | 0% | 1,848 | 931 | -50% | 0 | 0 | — |
case-12 | pass→pass | 15,666 | 6,518 | -58% | 1 | 1 | 0% | 1,243 | 1,425 | +15% | 0 | 0 | — |
case-13 | pass→fail | 16,707 | 5,475 | -67% | 1 | 1 | 0% | 2,422 | 946 | -61% | 0 | 0 | — |
case-14 | fail→fail | 8,842 | 14,064 | +59% | 1 | 1 | 0% | 1,450 | 904 | -38% | 0 | 0 | — |
case-15 | pass→fail | 12,859 | 5,768 | -55% | 1 | 1 | 0% | 2,118 | 967 | -54% | 0 | 0 | — |
case-16 | fail→fail | 23,321 | 20,156 | -14% | 1 | 1 | 0% | 2,610 | 1,018 | -61% | 0 | 0 | — |
case-17 | pass→fail | 13,668 | 2,969 | -78% | 1 | 1 | 0% | 2,171 | 1,122 | -48% | 0 | 0 | — |
case-18 | pass→fail | 11,515 | 4,532 | -61% | 1 | 1 | 0% | 1,742 | 939 | -46% | 0 | 0 | — |
case-19 | fail→fail | 15,242 | 7,745 | -49% | 1 | 1 | 0% | 2,298 | 995 | -57% | 0 | 0 | — |
case-20 | fail→fail | 13,562 | 5,165 | -62% | 1 | 1 | 0% | 2,077 | 1,009 | -51% | 0 | 0 | — |
case-21 | pass→fail | 11,998 | 4,862 | -59% | 1 | 1 | 0% | 1,832 | 1,031 | -44% | 0 | 0 | — |
case-22 | fail→fail | 17,132 | 4,653 | -73% | 1 | 1 | 0% | 2,825 | 965 | -66% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 5 counted toward the lift figure. The other 17 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of -100 percentage points is the difference between those two pass rates over the 5 comparable cases. 13 cases got worse with the skill loaded, and they are included in that figure.
Other measured skills in the registry, with their headline benchmark lift.