Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Activates a structured decorator framework that modifies how Claude reasons, responds, and formats output. Trigger this skill whenever a prompt contains one or more markers prefixed with +++, such as +++Reasoning, +++StepByStep, +++Tone(style=formal), or any other +++ marker, regardless of what follows in the prompt. Also trigger when the user asks to activate, list, or clear decorators, references the prompt decorator framework, or uses phrases like "use decorator", "apply +++", or "what decora
.claude/skills/bilal140202-prompt-decorators/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 29% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 108% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 192% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 61% | 0% |
Lightweight control markers that modify reasoning, style, structure, and session behavior. Activated by the +++ prefix. Multiple decorators can be combined in a single prompt. Each decorator's full definition lives in its own file under references/. Load only the file(s) you need. The Quick Reference table below maps every decorator to its file.
When a prompt contains +++ markers:
references/<DecoratorName>.md (e.g. +++Reasoning → references/Reasoning.md, +++StepByStep → references/StepByStep.md). Use the Quick Reference table in this file to map decorator names to file paths.+++Tone(style=formal), +++Refine(iterations=3)If an unrecognized decorator is used, note it and proceed with the rest.
Decorators operate in one of two scopes:
| Scope | Behavior | |-------|----------| | Message scope (default) | Decorator applies only to the current prompt | | Chat scope | Decorator persists across all subsequent messages until cleared |
+++ChatScope: switches all decorators in the same message to chat scope; they auto-apply to every future prompt+++MessageScope: pauses chat-scoped decorators for the current message only (does not erase them)+++Clear: removes all active chat-scoped decorators; +++Clear(+++Reasoning, +++Tone) removes only the named ones+++ActiveDecs: lists all currently active chat-scoped decorators+++AvailableDecs: lists all available decorators with name, description, and current statusTrack active chat-scoped decorators across the conversation and apply them automatically even when the user does not re-specify them.
All decorators in a prompt apply simultaneously and in order. They do not conflict, each governs a different dimension of the response (reasoning process, tone, format, verification, etc.). When decorators overlap in effect, use judgment to satisfy both.
Example:
+++Reasoning
+++StepByStep
+++Tone(style=formal)
Explain how transformer models handle attention.→ Begin with explicit reasoning, structure the body as labeled steps, maintain formal tone throughout.
| Category | Decorator | Definition File | |----------|-----------|----------------| | Reasoning & structure | +++Reasoning | references/Reasoning.md | | Reasoning & structure | +++StepByStep | references/StepByStep.md | | Reasoning & structure | +++Socratic | references/Socratic.md | | Reasoning & structure | +++Debate | references/Debate.md | | Evaluation & refinement | +++Critique | references/Critique.md | | Evaluation & refinement | +++Refine(iterations=N) | references/Refine.md | | Evaluation & refinement | +++Candor(level=L) | references/Candor.md | | Exploration & planning | +++Brainstorm | references/Brainstorm.md | | Exploration & planning | +++Planning | references/Planning.md | | Exploration & planning | +++Interactive | references/Interactive.md | | Exploration & planning | +++Import(topic=T) | references/Import.md | | Rewriting & clarification | +++Rewrite | references/Rewrite.md | | Output control | +++OutputFormat(format=F) | references/OutputFormat.md | | Output control | +++Tone(style=S) | references/Tone.md | | Verification | +++FactCheck | references/FactCheck.md | | Verification | +++CiteSources | references/CiteSources.md | | Verification | +++Validate | references/Validate.md | | Session management | +++ChatScope | references/ChatScope.md | | Session management | +++MessageScope | references/MessageScope.md | | Session management | +++Clear | references/Clear.md | | Session management | +++ActiveDecs | references/ActiveDecs.md | | Session management | +++AvailableDecs | references/AvailableDecs.md | | Export | +++Export(format=F) | references/Export.md | | Export | +++Dump | references/Dump.md |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 13,943 | 13,897 | -0% | 1 | 1 | 0% | 2,643 | 3,807 | +44% | 0 | 0 | — |
case-02 | pass→pass | 15,100 | 13,083 | -13% | 1 | 1 | 0% | 2,440 | 3,369 | +38% | 0 | 0 | — |
case-03 | fail→pass | 11,034 | 7,796 | -29% | 1 | 1 | 0% | 1,813 | 2,354 | +30% | 0 | 0 | — |
case-04 | pass→pass | 3,254 | 3,964 | +22% | 1 | 1 | 0% | 516 | 1,665 | +223% | 0 | 0 | — |
case-05 | fail→fail | 2,816 | 2,608 | -7% | 1 | 1 | 0% | 297 | 1,411 | +375% | 0 | 0 | — |
case-06 | pass→pass | 2,729 | 4,705 | +72% | 1 | 1 | 0% | 379 | 1,946 | +413% | 0 | 0 | — |
case-07 | fail→pass | 11,004 | 8,483 | -23% | 1 | 1 | 0% | 1,876 | 2,418 | +29% | 0 | 0 | — |
case-08 | fail→pass | 5,488 | 2,584 | -53% | 1 | 1 | 0% | 671 | 1,398 | +108% | 0 | 0 | — |
case-09 | fail→pass | 6,968 | 10,363 | +49% | 1 | 1 | 0% | 1,103 | 3,216 | +192% | 0 | 0 | — |
case-10 | pass→pass | 19,585 | 11,596 | -41% | 1 | 1 | 0% | 3,409 | 3,118 | -9% | 0 | 0 | — |
case-11 | fail→fail | 16,793 | 7,283 | -57% | 1 | 1 | 0% | 3,108 | 2,511 | -19% | 0 | 0 | — |
case-12 | fail→fail | 2,960 | 5,897 | +99% | 1 | 1 | 0% | 432 | 1,822 | +322% | 0 | 0 | — |
case-13 | fail→pass | 6,610 | 5,062 | -23% | 1 | 1 | 0% | 1,089 | 1,754 | +61% | 0 | 0 | — |
case-14 | fail→pass | 4,467 | 5,056 | +13% | 1 | 1 | 0% | 633 | 1,832 | +189% | 0 | 0 | — |
case-15 | fail→fail | 4,448 | 4,000 | -10% | 1 | 1 | 0% | 852 | 1,849 | +117% | 0 | 0 | — |
case-16 | fail→fail | 12,235 | 11,379 | -7% | 1 | 1 | 0% | 2,036 | 3,075 | +51% | 0 | 0 | — |
case-17 | fail→fail | 11,627 | 9,536 | -18% | 1 | 1 | 0% | 1,888 | 2,594 | +37% | 0 | 0 | — |
case-18 | fail→fail | 2,375 | 5,326 | +124% | 1 | 1 | 0% | 239 | 1,992 | +733% | 0 | 0 | — |
case-19 | fail→fail | 15,594 | 12,864 | -18% | 1 | 1 | 0% | 2,371 | 2,845 | +20% | 0 | 0 | — |
case-20 | pass→pass | 9,711 | 7,900 | -19% | 1 | 1 | 0% | 1,938 | 2,573 | +33% | 0 | 0 | — |
case-21 | fail→fail | 4,587 | 5,160 | +12% | 1 | 1 | 0% | 815 | 2,053 | +152% | 0 | 0 | — |
case-22 | pass→pass | 9,575 | 5,549 | -42% | 1 | 1 | 0% | 1,758 | 2,164 | +23% | 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 +27 percentage points is the difference between those two pass rates over the 22 comparable cases.
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.