Prompt decorators
Improve AI agents with reusable skills for structured workflows, better problem solving, and more consistent results
npx -y skills add Emagi6395/skills --skill prompt-decoratorsAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
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 decorators are available". This skill must be active for any +++ marker to take effect.
SKILL.md
4.8 KB, as published. Nobody here has run it
Prompt Decorators
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.
Parsing Decorators from a Prompt
When a prompt contains +++ markers:
- Extract all decorator names and parameters from the prompt (they appear before the main content)
- For each decorator, load its definition from
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. - Apply all active decorators simultaneously to the response
- Parameters are passed in parentheses, for example:
+++Tone(style=formal),+++Refine(iterations=3)
If an unrecognized decorator is used, note it and proceed with the rest.
Scope Rules
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 status
Track active chat-scoped decorators across the conversation and apply them automatically even when the user does not re-specify them.
Combining Decorators
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.
Quick Reference
| 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 |
Gives 0 of the 12 instructions most prompt engineering skills give
Counted across 563 of the 626 authors here whose files we hold, read 2026-08-06
- ask at most three clarifying questionsin 22 of 563, across 15 files
- respond in the user input languagein 14 of 563, across 9 files
- preserve the original intentin 13 of 563, across 11 files
- Establish baseline metrics and collect representative examplesin 12 of 563, across 2 files
- Identify failure modes and prioritize high-impact fixesin 12 of 563, across 2 files
- Apply prompt and workflow improvements with measurable goalsin 12 of 563, across 2 files
- Roll back quickly if quality or safety metrics regressin 12 of 563, across 2 files
- validate changes with tests and roll out in controlled stagesin 12 of 563, across 2 files
- generate quantitative baseline performance reportsin 12 of 563, across 2 files
- create representative test scenariosin 12 of 563, across 2 files
- treat prompts as codein 12 of 563, across 5 files
- test prompts on diverse inputsin 12 of 563, across 8 files
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once.