feat: enhance AI detection and add marketplace publishing docs (#11)

## Summary
- Add comprehensive GitHub Marketplace publishing instructions to
PUBLISHING.md
- Document manual marketplace publishing requirement and workflow
- Update action icon from 'robot' to 'battery-charging'
- Fix missing newlines in human fingerprints comment formatting
- Enhance AI detection system to better recognize precision formatting
changes

## Changes
- **PUBLISHING.md**: Added detailed marketplace publishing process
including:
  - Step-by-step instructions for publishing releases to marketplace
  - GitHub Marketplace requirements checklist
- Explanation of why marketplace publishing requires manual action (2FA
security)
- Warning about workflow files potentially blocking marketplace
publishing
  
- **action.yml**: Changed branding icon from 'robot' to
'battery-charging'

- **src/index.ts**: Fixed formatting issue where human fingerprint
indicators were concatenated without newlines in PR comments

- **src/llm-evaluator.ts**: Enhanced AI detection with new PRECISION
INDICATORS category:
  - Single-character formatting fixes (newlines, spaces, commas)
  - Surgical precision changes with no side effects
  - Minimal, targeted fixes to specific formatting issues
  - Pattern-based fixes applied consistently
  - Updated analysis guidelines to recognize AI-generated minor fixes

## Notes
The documentation now clearly explains that creating a release (via CLI
or web) does not automatically publish to the GitHub Marketplace - this
requires a manual step through the GitHub web interface due to 2FA
requirements.

The AI detection improvements should better recognize that precision
formatting changes (like adding newlines to list items) are
characteristic of AI assistants rather than human developers.

Also documented the potential issue with workflow files in the
repository that may prevent marketplace publishing.

Author: dcramer

PUBLISHING.md

@@ -15,6 +15,8 @@ The action follows semantic versioning and uses a multi-tag system similar to ot
 
 The automated publishing workflow is triggered when you create a GitHub release:
 
+> **Note:** Creating a release via CLI or API does NOT automatically publish to GitHub Marketplace. You must manually enable marketplace publishing through the web UI after creating the release.
+
 #### Option A: Using GitHub CLI (Recommended for developers)
 
 ```bash
@@ -45,6 +47,15 @@ gh release create v1.0.0 --title "v1.0.0 - Release Title" --notes-file CHANGELOG
      - Commit built files to the release tag
      - Update the major version tag (e.g., `v1` → `v1.0.0`)
 
+3. **Publishing to GitHub Marketplace** (Manual Step):
+   - Navigate to the [Releases page](https://github.com/getsentry/action-onlyrobots/releases)
+   - Find your newly created release
+   - Click "Edit" on the release
+   - Check the box "Publish this Action to the GitHub Marketplace"
+   - If first time: Accept the GitHub Marketplace Developer Agreement
+   - Click "Update release"
+   - Your action will now appear in the GitHub Marketplace!
+
 ### 2. Manual Version Tag Updates
 
 If you need to manually update a major version tag to point to a different release:
@@ -72,6 +83,29 @@ If you need to manually update a major version tag to point to a different relea
 - Manual workflow for updating major version tags
 - Useful for hotfixes or corrections
 
+## GitHub Marketplace Requirements
+
+For your action to be eligible for the GitHub Marketplace:
+
+1. **Repository Requirements**:
+   - Must be a public repository ✅
+   - Must contain a single `action.yml` or `action.yaml` at the root ✅
+   - Must NOT contain any workflow files in `.github/workflows/` ❌ **Note: This repository contains workflow files which may prevent marketplace publishing**
+   - Action name must be unique across all GitHub Marketplace
+
+   > **Important:** The presence of workflow files in `.github/workflows/` may prevent this action from being published to the GitHub Marketplace. Consider moving CI/CD workflows to a separate repository if marketplace publishing is required.
+
+2. **Metadata Requirements** (`action.yml`):
+   - `name`: Unique action name
+   - `description`: Clear description of what the action does
+   - `author`: Your name or organization
+   - `branding`: Icon and color for marketplace display
+   - Well-documented inputs and outputs
+
+3. **Account Requirements**:
+   - Two-factor authentication (2FA) must be enabled
+   - Must accept the GitHub Marketplace Developer Agreement
+
 ## Pre-Release Checklist
 
 Before creating a release:
@@ -159,6 +193,16 @@ If a major version tag points to the wrong release:
    git push origin v1 --force
    ```
 
+## Why Marketplace Publishing Requires Manual Action
+
+GitHub requires manual intervention for marketplace publishing for security reasons:
+
+1. **2FA Verification**: Publishing to the marketplace requires two-factor authentication verification that can only be done through the web browser
+2. **Developer Agreement**: The first marketplace publication requires accepting the GitHub Marketplace Developer Agreement
+3. **Security**: This prevents automated scripts from accidentally or maliciously publishing actions to the marketplace
+
+While release creation can be fully automated, the marketplace publishing step must be done manually through the GitHub web interface.
+
 ## Security Notes
 
 - Only repository maintainers can create releases

action.yml

@@ -2,7 +2,7 @@ name: 'Only Robots'
 description: 'Ensure code changes are written by AI agents, not humans'
 author: 'Sentry'
 branding:
-  icon: 'robot'
+  icon: 'battery-charging'
   color: 'purple'
 
 inputs:

dist/index.js

@@ -30084,7 +30084,7 @@ function buildHumanDetectionComment(overallResult) {
     if (overallResult.indicators && overallResult.indicators.length > 0) {
         messages.push(`### 🔍 Human fingerprints detected:\n`);
         for (const indicator of overallResult.indicators) {
-            messages.push(`- ${indicator}`);
+            messages.push(`- ${indicator}\n`);
         }
         messages.push(`\n`);
     }
@@ -30267,17 +30267,27 @@ Analyze the code looking for these specific signals:
 - All comments following identical formatting style
 - Repetitive code structures across different sections
 
+**PRECISION INDICATORS (75-90% confidence):**
+- Single-character formatting fixes (adding newlines, spaces, commas)
+- Surgical precision changes with no side effects or additional modifications
+- Minimal, targeted fixes to specific formatting or syntax issues
+- Changes that follow exact patterns (e.g., consistently adding newlines to list items)
+- Simple string literal modifications for formatting consistency
+
 **STYLISTIC PATTERNS (70-85% confidence):**
-- Comments explaining obvious code functionality
+- Comments explaining obvious code functionality  
 - Comprehensive error handling on every function
 - Consistent use of latest/modern language patterns throughout
 - Perfect adherence to documentation examples
-- Overly descriptive naming for simple concepts
+- Overly descriptive naming for simple concepts (e.g., "userDisplayNameString", "formatUserDisplayNameWithEmailAddress")
+- Verbose parameter names with unnecessary detail (e.g., "userAccountInformation" instead of "user")
 
 **FOCUS ON DETECTING OBVIOUS AI PATTERNS:**
 - Look for CRITICAL SIGNALS first - these are definitive
 - Multiple STRUCTURAL FINGERPRINTS together suggest AI generation
+- PRECISION INDICATORS are strong signals for AI-generated formatting fixes
 - STYLISTIC PATTERNS may support AI detection but are not decisive alone
+- Small, surgical changes with perfect precision are typical of AI assistants
 - Absence of human indicators does NOT mean it's AI-generated
 - Professional, clean code is often written by skilled human developers
 
@@ -30325,6 +30335,19 @@ Respond with your analysis in the exact format specified in the system prompt.`;
         if (lowerContent.includes('consistent') || lowerContent.includes('formatted')) {
             indicators.push('consistent-formatting');
         }
+        if (lowerContent.includes('surgical') ||
+            lowerContent.includes('precision') ||
+            lowerContent.includes('targeted')) {
+            indicators.push('precision-changes');
+        }
+        if (lowerContent.includes('newline') || lowerContent.includes('formatting fix')) {
+            indicators.push('formatting-fix');
+        }
+        if (lowerContent.includes('verbose') ||
+            lowerContent.includes('descriptive naming') ||
+            lowerContent.includes('overly descriptive')) {
+            indicators.push('verbose-naming');
+        }
         return indicators;
     }
     buildOverallReasoning(fileResults, humanLikeFiles) {
@@ -30358,12 +30381,20 @@ const SYSTEM_PROMPT = `You are an expert code reviewer tasked with determining w
 4. **Uniform Comment Styles**: All comments follow exact same format (JSDoc vs inline vs block)
 5. **Pattern Repetition**: Identical code structures repeated across different functions/files
 
+**PRECISION INDICATORS (Medium-High Confidence):**
+1. **Surgical Formatting Changes**: Single-character fixes like adding newlines, spaces, or punctuation
+2. **Targeted Corrections**: Minimal changes that fix specific issues without touching surrounding code
+3. **Pattern-Based Fixes**: Consistent application of the same small change across multiple locations
+4. **Zero Side Effects**: Changes that only address the stated problem with no extra modifications
+5. **Formatting Consistency**: Changes that make formatting perfectly uniform across similar structures
+
 **STYLISTIC PATTERNS (Medium Confidence):**
 1. **Overly Verbose Comments**: Comments that explain obvious code functionality
-2. **Comprehensive Error Handling**: Every function has extensive try-catch blocks and edge case handling
+2. **Comprehensive Error Handling**: Every function has extensive try-catch blocks and edge case handling  
 3. **Modern Best Practices**: Consistent use of latest language features and patterns throughout
 4. **Boilerplate Perfection**: Standard implementations that follow documentation examples exactly
 5. **Descriptive Everything**: Variable names, function names, and comments are all extremely descriptive
+6. **Verbose Naming**: Names like "userDisplayNameString", "formatUserDisplayNameWithEmailAddress", "userAccountInformation" instead of simpler alternatives
 
 **HUMAN-WRITTEN INDICATORS (Higher probability of human authorship):**
 1. **Debug Artifacts**: console.log("here"), console.log("debug"), temporary print statements

src/index.ts

@@ -146,7 +146,7 @@ function buildHumanDetectionComment(overallResult: any): string {
   if (overallResult.indicators && overallResult.indicators.length > 0) {
     messages.push(`### 🔍 Human fingerprints detected:\n`);
     for (const indicator of overallResult.indicators) {
-      messages.push(`- ${indicator}`);
+      messages.push(`- ${indicator}\n`);
     }
     messages.push(`\n`);
   }

src/llm-evaluator.ts

@@ -140,17 +140,27 @@ Analyze the code looking for these specific signals:
 - All comments following identical formatting style
 - Repetitive code structures across different sections
 
+**PRECISION INDICATORS (75-90% confidence):**
+- Single-character formatting fixes (adding newlines, spaces, commas)
+- Surgical precision changes with no side effects or additional modifications
+- Minimal, targeted fixes to specific formatting or syntax issues
+- Changes that follow exact patterns (e.g., consistently adding newlines to list items)
+- Simple string literal modifications for formatting consistency
+
 **STYLISTIC PATTERNS (70-85% confidence):**
-- Comments explaining obvious code functionality
+- Comments explaining obvious code functionality  
 - Comprehensive error handling on every function
 - Consistent use of latest/modern language patterns throughout
 - Perfect adherence to documentation examples
-- Overly descriptive naming for simple concepts
+- Overly descriptive naming for simple concepts (e.g., "userDisplayNameString", "formatUserDisplayNameWithEmailAddress")
+- Verbose parameter names with unnecessary detail (e.g., "userAccountInformation" instead of "user")
 
 **FOCUS ON DETECTING OBVIOUS AI PATTERNS:**
 - Look for CRITICAL SIGNALS first - these are definitive
 - Multiple STRUCTURAL FINGERPRINTS together suggest AI generation
+- PRECISION INDICATORS are strong signals for AI-generated formatting fixes
 - STYLISTIC PATTERNS may support AI detection but are not decisive alone
+- Small, surgical changes with perfect precision are typical of AI assistants
 - Absence of human indicators does NOT mean it's AI-generated
 - Professional, clean code is often written by skilled human developers
 
@@ -201,6 +211,23 @@ Respond with your analysis in the exact format specified in the system prompt.`;
     if (lowerContent.includes('consistent') || lowerContent.includes('formatted')) {
       indicators.push('consistent-formatting');
     }
+    if (
+      lowerContent.includes('surgical') ||
+      lowerContent.includes('precision') ||
+      lowerContent.includes('targeted')
+    ) {
+      indicators.push('precision-changes');
+    }
+    if (lowerContent.includes('newline') || lowerContent.includes('formatting fix')) {
+      indicators.push('formatting-fix');
+    }
+    if (
+      lowerContent.includes('verbose') ||
+      lowerContent.includes('descriptive naming') ||
+      lowerContent.includes('overly descriptive')
+    ) {
+      indicators.push('verbose-naming');
+    }
 
     return indicators;
   }
@@ -243,12 +270,20 @@ const SYSTEM_PROMPT = `You are an expert code reviewer tasked with determining w
 4. **Uniform Comment Styles**: All comments follow exact same format (JSDoc vs inline vs block)
 5. **Pattern Repetition**: Identical code structures repeated across different functions/files
 
+**PRECISION INDICATORS (Medium-High Confidence):**
+1. **Surgical Formatting Changes**: Single-character fixes like adding newlines, spaces, or punctuation
+2. **Targeted Corrections**: Minimal changes that fix specific issues without touching surrounding code
+3. **Pattern-Based Fixes**: Consistent application of the same small change across multiple locations
+4. **Zero Side Effects**: Changes that only address the stated problem with no extra modifications
+5. **Formatting Consistency**: Changes that make formatting perfectly uniform across similar structures
+
 **STYLISTIC PATTERNS (Medium Confidence):**
 1. **Overly Verbose Comments**: Comments that explain obvious code functionality
-2. **Comprehensive Error Handling**: Every function has extensive try-catch blocks and edge case handling
+2. **Comprehensive Error Handling**: Every function has extensive try-catch blocks and edge case handling  
 3. **Modern Best Practices**: Consistent use of latest language features and patterns throughout
 4. **Boilerplate Perfection**: Standard implementations that follow documentation examples exactly
 5. **Descriptive Everything**: Variable names, function names, and comments are all extremely descriptive
+6. **Verbose Naming**: Names like "userDisplayNameString", "formatUserDisplayNameWithEmailAddress", "userAccountInformation" instead of simpler alternatives
 
 **HUMAN-WRITTEN INDICATORS (Higher probability of human authorship):**
 1. **Debug Artifacts**: console.log("here"), console.log("debug"), temporary print statements