---

name: analyze-ambiguities description: Read a specification document and identify all ambiguities, gaps, unclear items, and implicit assumptions that need clarification or research. context: fork allowed-tools: Read user-invocable: false agent: spec-critic

Analyze Ambiguities - Specification Gap Detector

You are analyzing the specification for gaps and ambiguities.

Input: Path to spec document provided in $ARGUMENTS

Step 1: Read the Specification

Read the full specification document at the provided path.

Parse the YAML frontmatter to understand:

• Current phase (what-auto or how-auto)
• Current iteration count
• Previous convergence metrics

Step 2: Analyze for Ambiguities

Examine each section systematically:

In High Confidence Section

Look for:

• Items that reference undefined terms
• Statements that could be interpreted multiple ways
• Missing details (what, when, how, who)
• Implicit assumptions not stated explicitly
• Dependencies on external systems not specified

In Medium Confidence Section

Look for:

• Items that need verification against codebase
• Assumptions that should be promoted to high confidence or demoted to questions
• Vague language ("should", "might", "could", "probably")
• Missing acceptance criteria

In Open Questions Section

Look for:

• Questions that can be split into more specific sub-questions
• Questions that might already be answered elsewhere in spec (duplicates)
• Questions that are too broad to research effectively

Cross-Section Analysis

Look for:

• Contradictions between sections
• Missing connections (A depends on B but B not mentioned)
• Scope creep indicators
• Security, performance, error handling gaps

Step 3: Categorize Findings

For each ambiguity found, categorize as:

1. RESEARCH_NEEDED: Can likely be answered by searching codebase

   • Pattern: "How does existing feature X handle Y?"
   • Pattern: "What convention does project use for Z?"

2. HUMAN_NEEDED: Requires human decision or clarification

   • Pattern: "Should feature support X or Y approach?"
   • Pattern: "What is acceptable latency for this operation?"

3. DERIVABLE: Can be inferred from other spec items

   • Pattern: "If A is true, then B must be..."

4. OUT_OF_SCOPE: Should be explicitly marked as not included

   • Pattern: "Feature X mentioned but not core to this spec"

Step 4: Generate Search Strategies

For each RESEARCH_NEEDED item, suggest specific search strategies:

CopyAmbiguity: How does the project handle authentication tokens?
Search Strategy:
  - Glob: **/*auth*.{php,js,ts}
  - Grep: "token" in session-related files
  - Grep: "Bearer" or "JWT" patterns
  - Look for: middleware, session handlers

Step 5: Output Structured Report

Return findings in this format:

CopyAMBIGUITY ANALYSIS COMPLETE
Spec: {path}
Phase: {what-auto|how-auto}
Iteration: {n}

## Research Needed ({count})

### 1. {Brief description}
- Location: {section where found}
- Quote: "{relevant text from spec}"
- Search Strategy:
  - {specific search approach}
  - {alternative approach}

### 2. {Next ambiguity}
...

## Human Decision Needed ({count})

### 1. {Brief description}
- Location: {section}
- Options: {if identifiable}
- Context: {why this matters}

## Derivable ({count})

### 1. {Brief description}
- Can be derived from: {other spec items}
- Suggested resolution: {inference}

## Out of Scope ({count})

### 1. {Brief description}
- Reason: {why it's out of scope}
- Recommendation: {add to exclusions section}

## Summary
- Total ambiguities: {n}
- Research targets: {n}
- Awaiting human: {n}
- Quick fixes: {n}

Important Notes

• Be thorough but not pedantic - focus on ambiguities that matter
• An item in High Confidence that has gaps should be flagged
• Don't flag stylistic issues, only substantive gaps
• Prioritize ambiguities that block other understanding