# Shynlee04 Idumb V2 Intent Clarification > Skill: Intent Clarification - Skill: `tomevault-io/shynlee04-idumb-v2-intent-clarification` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/shynlee04-idumb-v2-intent-clarification` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/shynlee04-idumb-v2-intent-clarification/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/shynlee04-idumb-v2-intent-clarification --- # Skill: Intent Clarification Transform vague, underspecified research requests into precise, actionable research plans before executing any research activities. ## When to Use This Skill **Trigger this skill when:** - User says "research", "investigate", "analyze", "look into", "find out" without specifics - Request lacks: scope, goals, deliverables, success criteria, time constraints - Request is ambiguous about: what to research, why it's needed, who will use findings - User's prompt is > 10 lines or response is absurdly short → MUST trigger clarification - Request mixes research with implementation → separate concerns **Do NOT use when:** - Request is already specific: "Investigate why user login fails with error 403" - Request clearly defines: scope, goals, deliverables, success criteria - User explicitly provides context, constraints, and expected outcomes ## Intent Clarification Framework ### Step 1: Detect Ambiguity Check if the request is missing any of these critical elements: | Element | What It Looks Like | Why It Matters | |---------|-------------------|----------------| | **Scope** | "What exactly should I research?" | Avoids over-researching or missing key areas | | **Goal** | "What problem are we solving?" | Focuses research on actionable outcomes | | **Deliverables** | "What output do you expect?" | Defines success criteria upfront | | **Context** | "Background, constraints, assumptions?" | Prevents researching irrelevant areas | | **Audience** | "Who will use these findings?" | Determines depth and format of output | ### Step 2: Ask Targeted Questions Use the **5W1H framework** to structure clarification questions: ``` WHAT: What specifically do you want to know? WHY: Why is this research needed now? What will it enable? WHO: Who will use the findings? Decision-maker? Developer? Architect? WHERE: In what context will this be applied? Specific feature, system, or project? WHEN: By when do you need this? Any time constraints or deadlines? HOW: How detailed should the findings be? High-level summary vs deep dive? ``` **Question phrasing principles:** - Ask ONE question at a time (avoid overwhelming user) - Use binary choices when possible (A vs B) to reduce cognitive load - Provide examples to clarify abstract concepts - Offer default assumptions if user is uncertain ### Step 3: Validate with Examples After gathering answers, validate by paraphrasing back: ``` Let me confirm I understand correctly: **Research Scope:** [restate scope] **Primary Goal:** [restate goal] **Key Deliverables:** [list expected outputs] **Assumptions:** [list any assumptions] **Expected Format:** [bullet points, report, code examples?] **Success Criteria:** [how will you know I succeeded?] Is this accurate? Any adjustments needed before I begin research? ``` ### Step 4: Define Research Boundaries Explicitly state what will NOT be researched: ``` **Out of Scope:** - [Area X] - Not relevant to your goal - [Area Y] - Requires access I don't have - [Area Z] - Too time-consuming for current timeline **In Scope:** - [Primary research area] - Core focus - [Secondary area] - If time permits **Approach:** 1. [First step] 2. [Second step] 3. [Final validation] Ready to proceed? ``` ### Step 5: Establish Checkpoints Define validation moments during research: ``` **Research Checkpoints:** - ✓ After [X], I'll share initial findings - ✓ After [Y], I'll validate direction with you - ✓ Final deliverable: [what you'll receive] **Stop Conditions:** - If I find [condition], I'll pause for your input - If approach isn't working, I'll propose alternatives Shall I begin? ``` ## Common Ambiguity Patterns ### Pattern 1: Overly Broad Requests ``` User: "Research the codebase" ❌ Too vague Clarification: "I'd love to help research the codebase! To focus my research effectively: 1. What aspect of the codebase interests you? - Architecture patterns? - Performance bottlenecks? - Security vulnerabilities? - Tech stack decisions? 2. What's your goal with this research? - Understand how to contribute? - Prepare for a migration? - Audit code quality? - Something else? 3. What format would be most useful? - High-level overview? - Deep dive into specific areas? - Diagrams and documentation? - Code examples?" ``` ### Pattern 2: Context-Loss from Compaction ``` User: "Why is this slow?" (after context compaction) ❌ Missing context (what is "this"?) Clarification: "I notice we've lost some context from the conversation. To help you: 1. What specific functionality is slow? - API endpoint? UI component? Database query? Build process? 2. What does "slow" mean here? - Response time > X seconds? - Slow relative to what baseline? - User-reported latency? 3. When did this start happening? - After recent changes? - Always been slow? - Intermittent issue? Can you provide more context so I can investigate effectively?" ``` ### Pattern 3: Mixed Research and Implementation ``` User: "Research how to add OAuth and implement it" ❌ Two distinct concerns mixed together Clarification: "I see two distinct activities here. Let's separate them: **Research Phase:** - Investigate OAuth integration options - Analyze compatibility with your stack - Document recommended approach **Implementation Phase:** - Code the OAuth integration - Write tests - Update documentation Should I: 1. Start with research and get your approval before implementation? 2. Focus only on research right now (you'll handle implementation)? 3. Something else? What's your preference?" ``` ### Pattern 4: Missing Success Criteria ``` User: "Analyze the performance" ❌ No definition of what "good performance" means Clarification: "To analyze performance effectively, I need to know what "good" looks like: 1. What are your performance targets? - Response time < X milliseconds? - Throughput > Y requests/second? - Memory usage < Z MB? - Some other metric? 2. What's the baseline you're comparing against? - Current performance numbers? - Industry benchmarks? - Previous version? 3. What's the business impact? - User experience degradation? - Resource costs increasing? - Something else? Once I understand your goals, I can focus the analysis on what matters most." ``` ## Research Intent Clarification Checklist Before starting any research, verify: - [ ] Scope clearly defined (what will/won't be researched) - [ ] Goal articulated (what problem will this solve) - [ ] Deliverables specified (what output will you receive) - [ ] Success criteria established (how will you know it's complete) - [ ] Context gathered (constraints, assumptions, audience) - [ ] Checkpoints defined (when will we validate progress) - [ ] Stop conditions identified (when will I pause for input) - [ ] Timeline understood (any deadlines or time constraints) - [ ] Research vs implementation separated (if applicable) - [ ] Paraphrased back for confirmation (user validated understanding) ## Research Readiness Levels **Level 1 (Not Ready):** Vague request, multiple ambiguities → **Action:** Trigger this skill, ask clarification questions **Level 2 (Partially Ready):** Some clarity, but missing key details → **Action:** Ask targeted questions to fill gaps **Level 3 (Ready):** Specific request with clear goals, scope, deliverables → **Action:** Paraphrase back for confirmation, then proceed with research **Level 4 (Over-Defined):** Extremely detailed request → **Action:** Validate if this level of detail is necessary or if it indicates micromanagement ## Integration with Other Skills This skill works well in sequence with: - **gap-analysis** - After clarifying intent, systematically identify knowledge gaps - **multi-aspect-assessment** - Evaluate request from multiple angles (technical, business, UX, etc.) - **research-workflow-planner** - Create structured research plan once intent is clear - **opencode-primitive-selector** - Choose appropriate OpenCode tools for research execution ## Common Mistakes to Avoid ❌ **Assuming** what user means instead of asking → Always verify, never assume ❌ **Asking multiple questions at once** → Ask one question at a time, get answer, then ask next ❌ **Skipping paraphrase-back step** → Always validate understanding before proceeding ❌ **Proceeding with research when ambiguity remains** → Better to spend 2 minutes clarifying than 2 hours researching wrong thing ❌ **Not defining what's out of scope** → Explicitly state what you won't research to set expectations ❌ **Forgetting checkpoints** → Define validation moments to catch misalignment early ## Examples of Good vs Bad Clarification **Bad Clarification:** ``` What do you want to research? Be more specific. ``` **Good Clarification:** ``` To help you effectively, I'd like to understand your research goals: 1. What specific area of authentication are you investigating? - Login flow security? - Token refresh mechanism? - Role-based access control? - Something else? 2. What problem are you trying to solve? - Security vulnerability? - Performance issue? - Integration complexity? - Another concern? Let's start with the first question and we'll build from there. ``` ## Outcome After using this skill: - Research intent is crystal clear - Scope, goals, deliverables, and success criteria defined - Research plan established with checkpoints - Ready to proceed with `gap-analysis`, `multi-aspect-assessment`, or `research-workflow-planner` **Never start research without clear intent.** This skill is your first line of defense against wasted effort and misaligned deliverables. --- > Converted and distributed by [TomeVault](https://tomevault.io/claim/shynlee04) — claim your Tome and manage your conversions.