Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →This skill should be used when generating spec artifacts (research.md, requirements.md, design.md, tasks.md), formatting agent output, structuring phase results, or when any Ralph agent needs guidance on concise, scannable output formatting. Applies to all Ralph spec phase agents.
.claude/skills/tzachbon-communication-style/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -60% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -71% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -54% | 0% |
Be extremely concise. Sacrifice grammar for concision.
| Instead of | Write | |------------|-------| | "The user will be able to..." | "User can..." | | "This component is responsible for..." | "Handles..." | | "In order to achieve this, we need to..." | "Requires:" | | "It should be noted that..." | (delete) |
Use:
Every output follows this order:
1. Brief overview (2-3 sentences MAX)
2. Main content (tables, bullets, diagrams)
3. Unresolved questions (if any)
4. Numbered action steps (ALWAYS LAST)Action steps appear last because terminal output is read bottom-up -- the most important content occupies the most visible position.
markdown## Next Steps 1. Create auth module at src/auth/ 2. Add JWT dependency 3. Implement login endpoint 4. Add tests
Before action steps, list unresolved questions:
markdown## Unresolved Questions - OAuth provider preference? (Google, GitHub, both) - Session duration requirement? - Rate limiting needed?
Catches ambiguities before they become bugs.
| Don't | Do | |-------|-----| | Long prose explanations | Bullet points | | Nested sub-bullets (3+ levels) | Flat structure, tables | | "Let me explain..." | (just explain) | | Repeating context | Reference by ID | | Hedging language | Direct statements |
references/examples.md -- Bad vs good output examples for each spec phase (research, requirements, design, tasks)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-10 | pass→pass | 20,656 | 6,605 | -68% | 1 | 1 | 0% | 3,437 | 1,576 | -54% | 0 | 0 | — |
case-01 | fail→fail | 20,400 | 6,930 | -66% | 1 | 1 | 0% | 3,204 | 1,565 | -51% | 0 | 0 | — |
case-02 | fail→pass | 21,287 | 7,256 | -66% | 1 | 1 | 0% | 3,448 | 1,713 | -50% | 0 | 0 | — |
case-03 | fail→pass | 18,268 | 7,049 | -61% | 1 | 1 | 0% | 3,209 | 1,617 | -50% | 0 | 0 | — |
case-04 | fail→pass | 24,055 | 7,136 | -70% | 1 | 1 | 0% | 3,911 | 1,572 | -60% | 0 | 0 | — |
case-05 | fail→pass | 31,149 | 6,042 | -81% | 1 | 1 | 0% | 5,150 | 1,512 | -71% | 0 | 0 | — |
case-06 | fail→pass | 20,074 | 7,146 | -64% | 1 | 1 | 0% | 3,723 | 1,706 | -54% | 0 | 0 | — |
case-07 | fail→fail | 28,625 | 9,814 | -66% | 1 | 1 | 0% | 4,711 | 2,255 | -52% | 0 | 0 | — |
case-08 | fail→pass | 23,323 | 6,654 | -71% | 1 | 1 | 0% | 3,966 | 1,449 | -63% | 0 | 0 | — |
case-09 | fail→pass | 19,355 | 5,854 | -70% | 1 | 1 | 0% | 3,306 | 1,284 | -61% | 0 | 0 | — |
case-11 | fail→pass | 17,215 | 7,020 | -59% | 1 | 1 | 0% | 2,793 | 1,564 | -44% | 0 | 0 | — |
case-12 | fail→pass | 18,789 | 6,206 | -67% | 1 | 1 | 0% | 3,085 | 1,559 | -49% | 0 | 0 | — |
case-13 | fail→pass | 20,299 | 5,673 | -72% | 1 | 1 | 0% | 3,582 | 1,416 | -60% | 0 | 0 | — |
case-14 | fail→pass | 23,913 | 5,820 | -76% | 1 | 1 | 0% | 4,034 | 1,388 | -66% | 0 | 0 | — |
case-15 | pass→pass | 25,045 | 5,033 | -80% | 1 | 1 | 0% | 4,446 | 1,171 | -74% | 0 | 0 | — |
case-16 | fail→pass | 19,424 | 6,925 | -64% | 1 | 1 | 0% | 3,401 | 1,686 | -50% | 0 | 0 | — |
case-17 | fail→pass | 20,199 | 7,607 | -62% | 1 | 1 | 0% | 3,495 | 1,649 | -53% | 0 | 0 | — |
case-18 | fail→pass | 20,327 | 7,237 | -64% | 1 | 1 | 0% | 3,470 | 1,712 | -51% | 0 | 0 | — |
case-19 | fail→pass | 21,929 | 5,677 | -74% | 1 | 1 | 0% | 3,689 | 1,329 | -64% | 0 | 0 | — |
case-20 | pass→pass | 13,005 | 9,144 | -30% | 1 | 1 | 0% | 1,975 | 1,904 | -4% | 0 | 0 | — |
case-21 | pass→fail | 15,869 | 6,442 | -59% | 1 | 1 | 0% | 2,341 | 1,511 | -35% | 0 | 0 | — |
case-22 | pass→pass | 13,836 | 7,614 | -45% | 1 | 1 | 0% | 2,142 | 1,761 | -18% | 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. The headline lift of +64 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.