Session Interaction Logger

Comprehensive, detailed logging of every AI-assisted interaction for audit trails, evaluation evidence, and project history documentation.

Overview

This skill enhances the existing session-logger hook to capture full interaction detail — not just events, but the complete content of prompts, AI reasoning, decisions made, files modified, commands executed, and test results. All logs are stored in logs/copilot/ as structured JSON for easy parsing and report generation.

When to Use This Skill

• •At the start of every coding session (automatic via hook)
• •After each user prompt to log the full request context
• •When making code changes to record what was modified and why
• •When running commands to capture inputs and outputs
• •When asked to export or generate an audit report
• •When preparing deliverables for AI-usage evaluation

Why This Skill Exists

This project is being evaluated on:

1. •Full project history with prompts, context, and chat histories used to steer the AI
2. •Robust automated tests (Unit and Integration) proving correctness and reliability
3. •Clear documentation of decisions, assumptions, and how output was refined

This skill provides artifact #1 — a complete, verifiable record of all AI interactions.

Log Files

All logs are written to logs/copilot/ (created automatically):

Log Schema

interactions.log

{
  "timestamp": "2026-02-14T10:30:00Z",
  "sessionId": "abc123",
  "sequenceNumber": 1,
  "type": "user-prompt | ai-response | clarification | refinement",
  "content": "Full text of the prompt or response",
  "context": {
    "activeFile": "src/main/java/com/addi/challenge/Main.java",
    "openFiles": ["..."],
    "taskInProgress": "Implement lead validation"
  },
  "metadata": {
    "tokensEstimate": 150,
    "toolsUsed": ["read_file", "create_file"],
    "filesReferenced": ["design-doc.md"]
  }
}

file-changes.log

{
  "timestamp": "2026-02-14T10:35:00Z",
  "sessionId": "abc123",
  "action": "create | modify | delete",
  "filePath": "src/main/java/com/addi/challenge/domain/Lead.java",
  "reason": "Implement Lead record per design doc section 3.2",
  "linesAdded": 25,
  "linesRemoved": 0,
  "diffSummary": "Created Lead record with fields: id, firstName, lastName, email, birthDate, nationalId"
}

commands.log

{
  "timestamp": "2026-02-14T10:40:00Z",
  "sessionId": "abc123",
  "command": "./gradlew test",
  "purpose": "Run unit tests after implementing Lead validation",
  "exitCode": 0,
  "outputSummary": "BUILD SUCCESSFUL - 12 tests passed",
  "duration": "3.2s"
}

decisions.log

{
  "timestamp": "2026-02-14T10:32:00Z",
  "sessionId": "abc123",
  "decision": "Use Java Records for domain model",
  "rationale": "Records provide immutability, auto-generated equals/hashCode/toString, and align with Java 25 best practices",
  "alternatives": ["Traditional POJO with Lombok", "Plain class with manual methods"],
  "context": "Design doc section 3.2 specifies immutable domain objects",
  "category": "architecture | design | implementation | testing | tooling"
}

Step-by-Step Workflows

Workflow 1: Log a Full Interaction

After every user prompt and your response, append to logs/copilot/interactions.log:

1. •Generate a sessionId (reuse from session start or generate UUID)
2. •Increment sequenceNumber for ordering
3. •Log the user's prompt with type: "user-prompt"
4. •Log your response summary with type: "ai-response"
5. •Include context: active file, task being worked on, tools used

Script: Use scripts/log-interaction.sh to append entries:

.github/skills/session-interaction-logger/scripts/log-interaction.sh \
  --type "user-prompt" \
  --content "Implement the Lead domain record" \
  --active-file "Main.java" \
  --session-id "$SESSION_ID"

Workflow 2: Log File Changes

Every time you create, modify, or delete a file:

1. •Record the file path, action type, and reason
2. •Summarize the diff (lines added/removed, key changes)
3. •Append to logs/copilot/file-changes.log

Workflow 3: Log Commands

Every time a terminal command is executed:

1. •Record the command, its purpose, and output summary
2. •Include exit code and duration
3. •Append to logs/copilot/commands.log

Workflow 4: Log Decisions

When making architectural or design decisions:

1. •State the decision clearly
2. •Document the rationale and alternatives considered
3. •Reference relevant context (design doc sections, requirements)
4. •Append to logs/copilot/decisions.log

Workflow 5: Export Session Report

When asked to export or generate a report, run:

.github/skills/session-interaction-logger/scripts/export-session-report.sh

This generates logs/copilot/session-report.md with:

• •Session timeline
• •All prompts and response summaries
• •Files created/modified with reasons
• •Commands executed with results
• •Decisions made with rationale
• •Test results summary

Integration with Existing Hooks

This skill enhances the existing session-logger hook scripts. The hooks capture events automatically; this skill adds content-level detail on top of those events.

Best Practices

Privacy & Security

• •logs/ is in .gitignore — logs never get committed by accident
• •No secrets, tokens, or credentials are logged
• •Prompt content is logged as-is; ensure no sensitive data in prompts
• •To disable: set SKIP_LOGGING=true environment variable

Quick Reference

# View latest interactions
tail -20 logs/copilot/interactions.log | jq .

# Count interactions in current session
wc -l logs/copilot/interactions.log

# View all decisions
cat logs/copilot/decisions.log | jq .

# Export full session report
.github/skills/session-interaction-logger/scripts/export-session-report.sh

# View file change history
cat logs/copilot/file-changes.log | jq '.filePath + " - " + .action + " - " + .reason'