analyze-sessions

Analyze recorded session data from ALL Claude Code projects to identify patterns and automation opportunities.

Arguments

• --project <name> — Analyze only sessions from the named project (matches the project field, which is the basename of the working directory)
• --targets-only — Exclude toolkit sessions; only analyze target projects

Data Location

Session logs are stored centrally in ~/.claude/logs/sessions.jsonl in JSONL format (one compact JSON object per line).

Each entry contains:

{"session_id":"...","timestamp":"2026-01-23T12:00:00Z","project":"project-name","cwd":"/path/to/project","transcript_path":"/path/to/transcript"}

The session-end-logger hook is installed at the user level (~/.claude/hooks/session-end-logger.sh) so it captures sessions from every project.

Analysis Process

Copy this checklist and track progress:

Analyze Sessions Progress:
- [ ] Step 1: Read session logs
- [ ] Step 2: Calculate statistics (grouped by project)
- [ ] Step 3: Identify target projects
- [ ] Step 4: Analyze transcripts
- [ ] Step 5: Find automation opportunities
- [ ] Step 6: Cross-project insights
- [ ] Step 7: Generate report

1. Read Session Logs

Read ~/.claude/logs/sessions.jsonl and parse all entries.

If --project <name> was given, filter to entries where project matches. If --targets-only was given, identify toolkit projects (see step 3) and exclude them.

If the file doesn't exist or is empty:

NO SESSION DATA
===============
No sessions have been logged yet.

Session logging is installed at the user level (~/.claude/hooks/session-end-logger.sh).
Run a few Claude Code sessions in any project, then run /analyze-sessions again.

If the hook is not installed, run /install-hooks and select "session-end-logger".

2. Calculate Statistics (Grouped by Project)

SESSION ANALYSIS
================

Total sessions: {N}
Date range: {earliest} to {latest}
Unique projects: {count}

Sessions per project:
- {project1}: {count} sessions ({first_date} – {last_date})
- {project2}: {count} sessions ({first_date} – {last_date})

3. Identify Target Projects

Scan ~/Projects/ for directories containing .claude/toolkit-version.json. These are toolkit target projects. All other projects are either the toolkit itself or standalone.

Project Classification:
  Toolkit: {toolkit project name}
  Target projects: {list of target project names}
  Other: {list of other project names}

4. Analyze Transcripts (If Available)

For each session with an accessible transcript_path:

1. Read the transcript file
2. Look for these patterns:

Manual Interventions:

• AskUserQuestion tool calls — Questions asked during sessions
• Text containing "manually", "by hand", "human must"
• Repeated similar questions across sessions

Common Blockers:

• Text containing "BLOCKED", "blocked", "waiting for"
• Error patterns that recur
• Missing credentials/env vars

Tool Usage Patterns:

• Count of each tool type used
• Sequences of tools that repeat

Workflow Friction:

• "cd" into directories (wrong directory issues)
• Repeated file lookups (missing context)
• Re-reading same files multiple times

5. Identify Automation Opportunities

Based on patterns found, identify:

High Impact (Repeated across 3+ sessions):

• Manual steps that could be automated with curl/Bash
• Questions that could be pre-answered in AGENTS.md
• Browser checks that could use Playwright

Medium Impact (Repeated 2 times):

• Verification steps that could be codified
• Setup steps that could be scripted

Low Impact (Single occurrence but notable):

• One-off manual interventions that might recur

6. Cross-Project Insights

Compare patterns across projects:

CROSS-PROJECT COMPARISON
========================

| Project | Sessions | Avg Questions/Session | Top Blocker | Autonomy Level |
|---------|----------|-----------------------|-------------|----------------|
| {proj}  | {N}      | {avg}                 | {blocker}   | {High/Med/Low} |

Autonomy Level = High (0-1 questions/session), Med (2-3), Low (4+)

Identify:

• Which projects need more AGENTS.md guidance (high question rate)
• Which projects have more context overflows (re-reading patterns)
• Common patterns that appear in multiple projects (shared automation opportunities)
• Skills from the toolkit that are underused in target projects

6b. Review Findings Before Writing

Before writing the report file, output a summary of key findings to the user:

SESSION ANALYSIS PREVIEW
========================
Sessions analyzed: {N}
Top patterns found: {count}

Key findings:
1. {finding 1}
2. {finding 2}
3. {finding 3}

Writing full report to .claude/logs/ANALYSIS_REPORT.md...

This gives the user visibility before the file is written.

7. Generate Report

Write analysis to .claude/logs/ANALYSIS_REPORT.md (in the toolkit):

# Session Analysis Report

Generated: {timestamp}
Sessions analyzed: {N}
Date range: {earliest} to {latest}

## Summary

- Total sessions: {N}
- Projects: {list}
- Common patterns identified: {count}

## Per-Project Breakdown

### {Project Name} ({N} sessions)

**Type:** Toolkit | Target | Other
**Date range:** {first} – {last}
**Key patterns:**
- {pattern 1}
- {pattern 2}

## Cross-Project Insights

### Autonomy Levels
| Project | Sessions | Questions/Session | Autonomy |
|---------|----------|-------------------|----------|
| {proj}  | {N}      | {avg}             | {level}  |

### Shared Patterns
- {pattern that appears across multiple projects}

## Automation Opportunities

### High Priority

#### 1. {Pattern Name}
**Occurrences:** {N} sessions
**Projects:** {list of projects where this appears}
**Pattern:** {description of what keeps happening}
**Suggested Automation:**
- {specific automation approach}
- {implementation hint}

#### 2. {Pattern Name}
...

### Medium Priority

#### 1. {Pattern Name}
...

### Low Priority

#### 1. {Pattern Name}
...

## Recommended Actions

1. **Create new skill:** {skill name} — {what it would automate}
2. **Add to AGENTS.md:** {guidance to add}
3. **Create hook:** {hook description}

## Raw Statistics

### Questions Asked (AskUserQuestion)
| Question Pattern | Count | Projects |
|-----------------|-------|----------|
| {pattern} | {N} | {list} |

### Tools Used
| Tool | Total Uses | Avg per Session |
|------|------------|-----------------|
| {tool} | {N} | {avg} |

### Blockers Encountered
| Blocker Type | Count | Resolution |
|--------------|-------|------------|
| {type} | {N} | {how resolved} |

8. Output Summary

After writing the report:

ANALYSIS COMPLETE
=================

Analyzed: {N} sessions across {M} projects
Report: .claude/logs/ANALYSIS_REPORT.md

Key Findings:
- {top finding 1}
- {top finding 2}
- {top finding 3}

Cross-Project:
- {cross-project insight 1}
- {cross-project insight 2}

Recommended next steps:
1. Review .claude/logs/ANALYSIS_REPORT.md for full details
2. Consider creating skills for high-priority patterns
3. Update AGENTS.md with guidance for common questions

Error Handling

If ~/.claude/logs/ directory doesn't exist:

• Create the directory structure: mkdir -p ~/.claude/logs
• Report that no sessions have been logged yet
• Suggest running /install-hooks to set up session logging

If session log file is malformed:

• Skip malformed lines and continue processing valid entries
• Report the count of skipped entries in the output
• Note: "X entries skipped due to malformed data"

If transcript files are inaccessible:

• Continue analysis without transcript data
• Mark transcript-dependent insights as "Limited (transcripts unavailable)"
• Report which transcripts couldn't be read

If no patterns are found:

• Report "No automation opportunities identified"
• Suggest that more sessions may be needed for meaningful patterns
• Provide guidance: "Run /analyze-sessions again after 5+ additional sessions"

If ANALYSIS_REPORT.md cannot be written:

• Output the report to the terminal instead
• Report the write failure with the specific error
• Suggest checking directory permissions

Limitations

• Transcript files may not be accessible (permissions, deleted)
• Analysis quality depends on session data available
• Some patterns require human judgment to identify
• Cross-project analysis requires the user-level hook to be installed