Overview

Transform vague, ambiguous, or unclear requests into optimized, well-structured prompts through systematic assessment, pattern detection, and framework application across any domain.

What It Helps You Do

Use this skill to clarify and structure unclear requests before attempting to execute them. It transforms vague goals into concrete, actionable prompts ready for execution.

Activate it with:

/accelint-prompt-manager
Phrases like "optimize my prompt" or "clarify this request"
Related requests about vague goals, undefined success criteria, or prompt framework selection

The skill helps you:

Convert ambiguous requests into prompts with measurable success criteria
Identify and resolve interpretation conflicts that would cause misaligned results
Select appropriate prompt frameworks (CO-STAR, RISEN, RODES) based on task type
Remove token-wasting patterns and model-specific anti-patterns
Calibrate prompts for specific execution contexts (Claude Code, API calls, system prompts)

Good to know: This skill optimizes prompts—it doesn't execute the tasks they describe. You receive a refined prompt ready to use, not a completed task.

When to Use

The skill triggers on these patterns:

Vague goals: "improve our X", "I don't know where to start", "kinda lost on what to do"
Undefined success: "should look professional", "explain clearly", missing constraints/audience/format
Ambiguous requirements: "analyze the data" (what to look for?), "make it robust" (how defined?)
Meta-prompting: "optimize this prompt", "review my instructions", learning about frameworks like CO-STAR/RISEN/RODES
Communication unclear: "how do I explain this", "help me figure out what to ask"

How It Works

The skill follows a four-phase workflow calibrated to your skill level (newcomer, intermediate, expert).

Phase 1: Intake & Assessment

Extracts core intent and assesses task complexity
Identifies execution context (interactive chat, API call, system prompt)
Recommends plan mode for complex tasks with 3+ interdependent decisions or 5+ sequential phases

Phase 2: Pattern Detection

Scans for credit-killing patterns (fabrication techniques, inappropriate Chain-of-Thought instructions, front-loaded context)
Flags ambiguities with 2-3 interpretation options and trade-off implications
Identifies missing context (audience, success criteria, constraints, format)

Phase 3: Framework Selection & Optimization

Applies the appropriate framework silently based on task type:

CO-STAR: Structured output needs (format-driven)
RISEN: Multi-step procedures (process-driven)
RODES: Clarification through examples (example-driven)

Uses safe optimization techniques:

Replaces vague terms with concrete criteria
Positions critical info at start/end (avoids lost-in-middle problem)
Eliminates ambiguous pronouns in multi-step instructions
Adds constraints to define creative boundaries

Phase 4: Validation & Handoff

Runs quality checks and flags remaining ambiguities needing your decision
Delivers optimized prompt directly in a markdown code block for easy copying
Offers to save to file and/or copy to clipboard
Provides iteration options for refinement

What You Get

An optimized prompt presented directly in your conversation, formatted in a markdown code block for easy copying.

For newcomers: before/after comparison with explanations. For experts: concise optimization notes.

Optional post-delivery: save to markdown file, copy to clipboard, or both.

Examples

Example: Clarifying a Vague Request

/accelint-prompt-manager "make our documentation better"

The skill turns the vague request into a concrete prompt with success criteria, audience, constraints, and output format.

Context: Technical documentation for a Node.js library README
Objective: Reduce new user onboarding time from 30 minutes to &lt;10 minutes
Success Criteria:
  - Quick Start completable in &lt;3 minutes
  - API reference with copy-paste examples for top 5 use cases
  - Troubleshooting section for 3 most common setup errors
Audience: JavaScript developers with npm experience, no prior library knowledge
Constraints:
  - Keep README under 500 lines
  - Include runnable code examples, not pseudocode
  - Link to advanced topics rather than embedding everything
Format: Markdown with section headers and table of contents

Example: Resolving Ambiguity

/accelint-prompt-manager "Analyze this codebase and tell me what's wrong"

The skill narrows the scope, defines priorities, and turns an open-ended request into a usable review prompt.

Role: Senior code reviewer conducting security and performance audit
Objective: Identify critical issues blocking production deployment in [codebase]
Focus Areas (prioritized):
  1. Security vulnerabilities (SQL injection, XSS, exposed secrets)
  2. Performance bottlenecks (O(n²) algorithms, N+1 queries, memory leaks)
  3. Error handling gaps (unhandled rejections, missing validation)
Scope: Scan src/ directory only, exclude test files and vendor code
Output Format:
  - Executive summary with risk levels (Critical/High/Medium/Low)
  - Per-issue: location, impact, remediation steps
  - Prioritized fix list with estimated effort
Constraints: Complete review within token budget; flag but don't fix issues

Good to Know

Good to know: The skill never fetches external resources or researches topics. If you provide a prompt mentioning "GitHub APIs", it optimizes the prompt text itself—it doesn't investigate those systems.

Good to know: Optimization strategies are model-specific. Claude 4.5+ uses extended thinking natively and doesn't benefit from explicit "think step by step" instructions, while older models might.

Good to know: The skill adapts communication style to your level. Newcomers receive explanations of detected patterns. Experts get pattern names and line numbers directly.

Overview

On this page