<?xml version="1.0" encoding="utf-8"?>
<entry xmlns="http://www.w3.org/2005/Atom" xml:base="https://docs.spec-kitty.ai/adr/2.x/2026-02-11-4-agent-attribution-tracking/">
  <id>https://docs.spec-kitty.ai/adr/2.x/2026-02-11-4-agent-attribution-tracking/</id>
  <title>Agent Attribution Tracking for Transparency and Compliance</title>
  <updated>2026-02-11T00:00:00.000Z</updated>
  <author><name>Spec Kitty</name></author>
  <link rel="alternate" type="text/html" href="https://docs.spec-kitty.ai/adr/2.x/2026-02-11-4-agent-attribution-tracking/"/>
  <summary>Line-level attribution joining git blame to agent execution logs, so regulated teams can say which model or which human authored any particular line of code.</summary>
  <content type="html">&lt;h2 id=&quot;context-and-problem-statement&quot;&gt;Context and Problem Statement&lt;/h2&gt;
&lt;p&gt;As AI agents write more code in production systems, enterprises need clear attribution tracking to answer critical questions:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Accountability&lt;/strong&gt;: Which lines were written by AI vs humans? (For code review, debugging, blame)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Compliance&lt;/strong&gt;: Audit trails for regulated industries (finance, healthcare, government) - “who wrote this code?”&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Quality&lt;/strong&gt;: Correlate agent-written code with bug rates, security vulnerabilities (learning which agents/prompts work best)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Transparency&lt;/strong&gt;: Build trust with users, stakeholders, regulators by showing clear AI contribution&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Policy enforcement&lt;/strong&gt;: Enable rules like “AI-generated code requires human review” (governance)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Licensing&lt;/strong&gt;: Legal questions emerging around AI-generated code ownership, copyright&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;Currently, Spec Kitty tracks WP-level metadata (which agent executed WP03), but NOT line-level attribution. Users cannot answer “which specific lines in main.py were written by Claude vs human edits?”&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Industry Context&lt;/strong&gt;:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Entire.io&lt;/strong&gt; tracks line-level attribution (agent vs human %) - competitive parity required&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;GitHub Copilot&lt;/strong&gt; adds comments marking AI-generated code - transparency precedent&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;The Register (2026-02-03)&lt;/strong&gt;: GitHub considering kill switches for low-quality AI PRs - attribution enables quality tracking&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Enterprise Need&lt;/strong&gt;: Fortune 500 customers evaluating Spec Kitty ask “Can we track which code was AI-generated?” for compliance, governance, learning.&lt;/p&gt;
&lt;h2 id=&quot;decision-drivers&quot;&gt;Decision Drivers&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Compliance&lt;/strong&gt;: Regulated industries (finance, healthcare) require audit trails showing who wrote each line&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Accountability&lt;/strong&gt;: Clear attribution enables code review (“review AI code more carefully”), debugging (“this bug came from agent X”)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Quality learning&lt;/strong&gt;: Measure which agents/prompts produce better code (iterate on prompts, agent selection)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Transparency&lt;/strong&gt;: Build trust with users, stakeholders, regulators&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Competitive parity&lt;/strong&gt;: Entire.io has line-level attribution - must match for Tier 1 threat response&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Policy enforcement&lt;/strong&gt;: Enable governance rules (“AI code requires review”, “limit AI to &amp;#x3C;50% of PR”)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Legal defensibility&lt;/strong&gt;: Address emerging questions on AI code ownership, copyright&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;considered-options&quot;&gt;Considered Options&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Option 1&lt;/strong&gt;: Line-level attribution (agent vs human, with agent ID) - chosen&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Option 2&lt;/strong&gt;: File-level attribution (too coarse)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Option 3&lt;/strong&gt;: Commit-level attribution (even coarser)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Option 4&lt;/strong&gt;: No attribution (status quo)&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;decision-outcome&quot;&gt;Decision Outcome&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Chosen option:&lt;/strong&gt; “Line-level attribution integrated with Git blame and agent execution logs”, because:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Industry standard&lt;/strong&gt;: Line-level is how developers think about code (Git blame operates at line level)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Compliance granularity&lt;/strong&gt;: Auditors want line-level detail (“show me every AI-generated line in this security function”)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Competitive parity&lt;/strong&gt;: Matches Entire.io’s attribution model (Tier 1 threat response)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Quality insights&lt;/strong&gt;: Enables detailed analysis (which agent wrote buggy line 47?)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Existing foundation&lt;/strong&gt;: Git blame provides baseline, Co-Authored-By markers identify agent commits&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Dashboard value&lt;/strong&gt;: Color-coded file views (green=human, blue=agent) provide instant visual clarity&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;&lt;strong&gt;Implementation&lt;/strong&gt;: Create &lt;code dir=&quot;auto&quot;&gt;AttributionTracker&lt;/code&gt; class in &lt;code dir=&quot;auto&quot;&gt;events/attribution.py&lt;/code&gt; that integrates Git blame (for baseline) with agent execution logs (for AI attribution) and Entire checkpoint data (if imported). Store attribution as structured metadata in event payloads (extends ADR 2026-02-09-4: Cross-Repo Evidence Completion).&lt;/p&gt;
&lt;h3 id=&quot;consequences&quot;&gt;Consequences&lt;/h3&gt;
&lt;h4 id=&quot;positive&quot;&gt;Positive&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Compliance&lt;/strong&gt;: Audit trails satisfy regulatory requirements (SOC2, HIPAA, financial regulations)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Accountability&lt;/strong&gt;: Clear blame for bugs (“line 47 was written by claude-sonnet-4 in WP03”)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Quality learning&lt;/strong&gt;: Correlate agents with quality metrics (bug rate, security vulnerabilities, review rejection)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Transparency&lt;/strong&gt;: Users see exactly what AI contributed (builds trust with teams, stakeholders, regulators)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Policy enforcement&lt;/strong&gt;: Enable governance (“AI code requires review”, “max 50% AI per file”)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Competitive parity&lt;/strong&gt;: Matches Entire.io’s attribution capability (reduces feature gap)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Dashboard value&lt;/strong&gt;: Visual attribution (color-coded files) makes AI contribution instantly visible&lt;/li&gt;
&lt;/ul&gt;
&lt;h4 id=&quot;negative&quot;&gt;Negative&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Storage overhead&lt;/strong&gt;: Attribution metadata per line adds storage (estimate: +10-20% event log size)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Performance cost&lt;/strong&gt;: Git blame is slow on large files (100ms-1s per file)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Maintenance&lt;/strong&gt;: Must keep attribution detection accurate (Co-Authored-By markers, agent commit detection)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Privacy concern&lt;/strong&gt;: Some users may not want line-level tracking (feels invasive)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Attribution ambiguity&lt;/strong&gt;: Human edits to AI-generated lines blur attribution (who owns line? agent or human?)&lt;/li&gt;
&lt;/ul&gt;
&lt;h4 id=&quot;neutral&quot;&gt;Neutral&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Opt-in for privacy&lt;/strong&gt;: Attribution tracking can be Tier 3+ (not required for local-only Tier 1 users)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Granularity trade-off&lt;/strong&gt;: Line-level is detailed but may be noisy (every whitespace change tracked)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Dashboard dependency&lt;/strong&gt;: Full value requires dashboard visualization (CLI shows text summary only)&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;confirmation&quot;&gt;Confirmation&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Success Metrics&lt;/strong&gt;:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Accuracy&lt;/strong&gt;: &gt;95% correct detection of agent vs human attribution (validated against known ground truth)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Compliance value&lt;/strong&gt;: 3+ enterprise customers cite attribution as key requirement (validate market need)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Adoption&lt;/strong&gt;: 50%+ of users enable attribution tracking (Tier 3+)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Performance&lt;/strong&gt;: Git blame &amp;#x3C;500ms p95 for typical files (&amp;#x3C;1000 lines)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Quality insights&lt;/strong&gt;: Identify correlation between agent type and bug rate (validates learning use case)&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Validation Timeline&lt;/strong&gt;:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Month 1&lt;/strong&gt;: Implement AttributionTracker, test with mixed agent/human repos&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Month 2&lt;/strong&gt;: Dashboard visualization (color-coded file views)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Month 3&lt;/strong&gt;: CLI command (&lt;code dir=&quot;auto&quot;&gt;spec-kitty attribution &amp;#x3C;file&gt;&lt;/code&gt;) for text summary&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Month 4-6&lt;/strong&gt;: Monitor adoption, accuracy, performance; iterate on detection logic&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Confidence Level&lt;/strong&gt;: &lt;strong&gt;HIGH&lt;/strong&gt; (8/10)&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Clear enterprise need (compliance, governance, transparency)&lt;/li&gt;
&lt;li&gt;Proven by Entire.io’s implementation (validates technical feasibility)&lt;/li&gt;
&lt;li&gt;Foundation exists (Git blame, Co-Authored-By markers)&lt;/li&gt;
&lt;li&gt;Main risk: Storage overhead (mitigable with compression, archival)&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;pros-and-cons-of-the-options&quot;&gt;Pros and Cons of the Options&lt;/h2&gt;
&lt;h3 id=&quot;line-level-attribution-chosen&quot;&gt;Line-Level Attribution (Chosen)&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Description&lt;/strong&gt;: Track attribution at line granularity (line 47 = claude-sonnet-4, line 48 = &lt;a href=&quot;mailto:user@example.com&quot;&gt;user@example.com&lt;/a&gt;).&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Industry standard&lt;/strong&gt;: Line-level matches developer mental model (Git blame)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Compliance granularity&lt;/strong&gt;: Satisfies audit requirements (“show me every AI line”)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Detailed insights&lt;/strong&gt;: Enables fine-grained quality analysis (which agent wrote buggy line?)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Dashboard value&lt;/strong&gt;: Color-coded visualization makes AI contribution instantly visible&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Competitive parity&lt;/strong&gt;: Matches Entire.io’s line-level attribution&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Existing integration&lt;/strong&gt;: Git blame provides baseline&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Cons:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Storage overhead&lt;/strong&gt;: Attribution per line adds 10-20% to event log&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Performance cost&lt;/strong&gt;: Git blame is slow (100ms-1s per file)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Attribution ambiguity&lt;/strong&gt;: Human edits to AI lines blur ownership&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Noise&lt;/strong&gt;: Every whitespace change tracked (may be too granular)&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;file-level-attribution&quot;&gt;File-Level Attribution&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Description&lt;/strong&gt;: Track attribution at file granularity (main.py = 70% agent, 30% human).&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Simpler&lt;/strong&gt;: Less storage, faster computation&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Aggregate view&lt;/strong&gt;: File-level percentages are useful for dashboards&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Cons:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Too coarse for compliance&lt;/strong&gt;: Auditors want line-level detail&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Lost insights&lt;/strong&gt;: Cannot identify which specific lines are buggy/secure&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Poor debugging&lt;/strong&gt;: File-level doesn’t help isolate bugs to specific lines&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Competitive gap&lt;/strong&gt;: Entire.io has line-level (we’d be behind)&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;commit-level-attribution&quot;&gt;Commit-Level Attribution&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Description&lt;/strong&gt;: Track attribution at commit granularity (commit abc123 = agent, commit def456 = human).&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Simplest&lt;/strong&gt;: Minimal storage, fast computation&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Existing metadata&lt;/strong&gt;: Git commits already have author info&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Cons:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Even coarser than file-level&lt;/strong&gt;: Commits span multiple files, many lines&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Useless for compliance&lt;/strong&gt;: Auditors need line-level, not commit-level&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;No quality insights&lt;/strong&gt;: Cannot correlate specific code with bugs&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Not competitive&lt;/strong&gt;: No competitor tracks only at commit level&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;no-attribution&quot;&gt;No Attribution&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Description&lt;/strong&gt;: Status quo - no line-level tracking.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Zero overhead&lt;/strong&gt;: No storage, no performance cost&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Simple&lt;/strong&gt;: No attribution code to maintain&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Cons:&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Compliance blocker&lt;/strong&gt;: Regulated industries cannot adopt Spec Kitty without attribution&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Quality blind spot&lt;/strong&gt;: Cannot learn which agents produce better code&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Competitive gap&lt;/strong&gt;: Entire.io has attribution, we don’t (feature gap)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Trust issue&lt;/strong&gt;: Users cannot verify AI contribution (transparency problem)&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;more-information&quot;&gt;More Information&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;References&lt;/strong&gt;:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Competitive analysis: &lt;code dir=&quot;auto&quot;&gt;competitive/tier-1-threats/entire-io/THREAT-ASSESSMENT.md&lt;/code&gt; (Entire’s attribution model)&lt;/li&gt;
&lt;li&gt;Entire.io codebase: &lt;a href=&quot;https://github.com/entireio/cli&quot;&gt;https://github.com/entireio/cli&lt;/a&gt; (see attribution tracking implementation)&lt;/li&gt;
&lt;li&gt;Product requirements: &lt;code dir=&quot;auto&quot;&gt;product-ideas/prd-agent-orchestration-integration-v1.md&lt;/code&gt; (AD-004)&lt;/li&gt;
&lt;li&gt;Integration spec: &lt;code dir=&quot;auto&quot;&gt;competitive/tier-1-threats/entire-io/INTEGRATION-SPEC.md&lt;/code&gt; (Section 1.2)&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Implementation Files&lt;/strong&gt;:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;events/attribution.py&lt;/code&gt; - AttributionTracker class&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;specify_cli/commands/attribution.py&lt;/code&gt; - CLI command for text summary&lt;/li&gt;
&lt;li&gt;Dashboard: Color-coded file views (green=human, blue=agent)&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Related ADRs&lt;/strong&gt;:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;ADR-2026-02-09-4: Cross-Repo Evidence Completion (attribution metadata extends evidence payloads)&lt;/li&gt;
&lt;li&gt;ADR-2026-02-11-3: Entire.io Checkpoint Import (Entire checkpoints provide attribution data)&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Detection Strategy&lt;/strong&gt;:&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Method 1: Git Blame + Co-Authored-By Markers&lt;/strong&gt; (primary)&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Run &lt;code dir=&quot;auto&quot;&gt;git blame --line-porcelain&lt;/code&gt; on file&lt;/li&gt;
&lt;li&gt;Check commit message for “Co-Authored-By: Claude” or “Co-Authored-By: Codex”&lt;/li&gt;
&lt;li&gt;Extract agent ID from Co-Authored-By (e.g., “Claude Sonnet 4.5” → “claude-sonnet-4.5”)&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Method 2: Agent Execution Logs&lt;/strong&gt; (secondary)&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Track file modifications in agent execution logs (tool calls, file writes)&lt;/li&gt;
&lt;li&gt;Cross-reference with Git commits to confirm attribution&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Method 3: Entire Checkpoint Import&lt;/strong&gt; (optional)&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;If user imports Entire checkpoints, extract attribution from file_changes diffs&lt;/li&gt;
&lt;li&gt;Merge with Git blame data for comprehensive view&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;CLI Usage Example&lt;/strong&gt;:&lt;/p&gt;
&lt;div class=&quot;expressive-code&quot;&gt;&lt;figure class=&quot;frame is-terminal not-content&quot;&gt;&lt;figcaption class=&quot;header&quot;&gt;&lt;span class=&quot;title&quot;&gt;&lt;/span&gt;&lt;span class=&quot;sr-only&quot;&gt;Terminal window&lt;/span&gt;&lt;/figcaption&gt;&lt;pre data-language=&quot;bash&quot;&gt;&lt;code&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;$&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;spec-kitty&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;attribution&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;src/main.py&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;
&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;Attribution&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;for&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;src/main.py&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;
&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;Total&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;lines:&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#F78C6C;--1:#AA0982&quot;&gt;245&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;Agent:&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#F78C6C;--1:#AA0982&quot;&gt;180&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; (73.5%)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;Human:&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#F78C6C;--1:#AA0982&quot;&gt;65&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; (26.5%)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;
&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;By&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;Agent:&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span class=&quot;indent&quot;&gt;  &lt;/span&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;claude-sonnet-4:&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#F78C6C;--1:#AA0982&quot;&gt;150&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;lines&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; (61.2%)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span class=&quot;indent&quot;&gt;  &lt;/span&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;codex-gpt4:&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#F78C6C;--1:#AA0982&quot;&gt;30&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;lines&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; (12.2%)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;
&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;By&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;Human:&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span class=&quot;indent&quot;&gt;  &lt;/span&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;user@example.com:&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#F78C6C;--1:#AA0982&quot;&gt;45&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;lines&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; (18.4%)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span class=&quot;indent&quot;&gt;  &lt;/span&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;reviewer@example.com:&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#F78C6C;--1:#AA0982&quot;&gt;20&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;lines&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; (8.2%)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;
&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;Line-by-line&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; (first &lt;/span&gt;&lt;span style=&quot;--0:#F78C6C;--1:#AA0982&quot;&gt;20&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;lines&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt;):&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span class=&quot;indent&quot;&gt;  &lt;/span&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;1-15:&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;agent&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; (claude-sonnet-4)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span class=&quot;indent&quot;&gt;  &lt;/span&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;16-20:&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;human&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; (user@example.com)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span class=&quot;indent&quot;&gt;  &lt;/span&gt;&lt;span style=&quot;--0:#82AAFF;--1:#3B61B0&quot;&gt;21-50:&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; &lt;/span&gt;&lt;span style=&quot;--0:#ECC48D;--1:#3B61B0&quot;&gt;agent&lt;/span&gt;&lt;span style=&quot;--0:#D6DEEB;--1:#403F53&quot;&gt; (claude-sonnet-4)&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;div class=&quot;ec-line&quot;&gt;&lt;div class=&quot;code&quot;&gt;&lt;span class=&quot;indent&quot;&gt;  &lt;/span&gt;&lt;span style=&quot;--0:#C5E478;--1:#3B61B0&quot;&gt;...&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/figure&gt;&lt;/div&gt;
&lt;p&gt;&lt;strong&gt;Dashboard Visualization&lt;/strong&gt;:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;File view: Color-coded lines (green = human, blue = agent, yellow = mixed)&lt;/li&gt;
&lt;li&gt;Hover tooltip: “Line 47: claude-sonnet-4 (2026-02-10 14:32)”&lt;/li&gt;
&lt;li&gt;File summary: “73.5% agent, 26.5% human”&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Privacy Tier Integration&lt;/strong&gt;:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Tier 1 (local-only)&lt;/strong&gt;: Attribution stored locally, never synced&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Tier 2 (workflow metadata)&lt;/strong&gt;: Attribution metadata synced (no code content)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Tier 3 (activity logs)&lt;/strong&gt;: Full attribution visible in dashboard&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Tier 4 (telemetry)&lt;/strong&gt;: Aggregate attribution stats (% AI code across all users)&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Compliance Use Cases&lt;/strong&gt;:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;SOC2&lt;/strong&gt;: Audit trail showing who modified security-sensitive code&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;HIPAA&lt;/strong&gt;: Track AI contribution to healthcare data processing logic&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Financial regulations&lt;/strong&gt;: Demonstrate compliance with code review policies&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Rollback Plan&lt;/strong&gt;:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;If storage overhead unacceptable: Make attribution opt-in (Tier 3+ only)&lt;/li&gt;
&lt;li&gt;If performance cost too high: Lazy attribution (compute on-demand, not stored)&lt;/li&gt;
&lt;li&gt;If privacy concerns: Allow disabling attribution tracking (configuration flag)&lt;/li&gt;
&lt;/ul&gt;</content>
</entry>
