Session End Skill

Automates the completion of a development session. Verifies tasks, runs quality checks, and updates documentation.

When to Use

Activate this skill when:

• User says "finish session", "complete session", or "end session"
• User says "session is done" or "ready to wrap up"
• User asks "what's left to finish the session?"
• User wants to wrap up session work

Project Configuration

This skill reads project-specific settings from .claude/project-config.md. If the config doesn't exist, the skill will prompt to create one.

Expected config sections:

## Quality Checks

Commands to run before completing a session:

### Format
```bash
cargo fmt --all

Build

cargo build --workspace

Lint

cargo clippy --workspace --all-targets

Test

cargo test --workspace

Custom Validation (Optional)

cargo run --bin validate_ron

The skill reads these commands and executes them in order.

## Instructions

### Step 0: Read Project Config

1. Check if `.claude/project-config.md` exists
2. If **YES**:
   - Parse `## Documentation Paths` section to get:
     - Development log path (default: `docs/development-log.md`)
     - Work streams path (default: `docs/work-streams/`)
     - Sessions path (default: `docs/sessions/`)
   - Parse `## Quality Checks` section for build commands
3. If **NO**: Prompt user to create one:

No project config found. Would you like to create one?

I need to know your project's quality check commands:

• Format command (e.g., "cargo fmt", "npm run format", "go fmt ./...")
• Build command (e.g., "cargo build", "npm run build", "go build ./...")
• Lint command (e.g., "cargo clippy", "npm run lint", "golangci-lint run")
• Test command (e.g., "cargo test", "npm test", "go test ./...")
• Any custom validation commands?

Create `.claude/project-config.md` with user's answers, then continue.

### Step 1: Find Current Session

**Identify the active session:**

1. Check TodoWrite list for session context (if available)
2. Glob all sessions: `[sessions-path]/*/*.md` (from config)
3. Filter and sort to find most recent with "Status: In Progress"
4. Read session file

**If no active session found:**
- Error: "No active session found. Use /session-start first?"
- List available in-progress sessions if any

### Step 2: Verify Task Completion

1. Read session log "Tasks" section
2. Read session log "What Was Accomplished" section
3. Compare tasks vs accomplishments:
- Are all tasks addressed?
- Are there incomplete tasks?
4. Check TodoWrite list (if available):
- Any todos still "pending" or "in_progress"?

5. If tasks incomplete:

Incomplete tasks detected:

• Task that's not done
• Another incomplete task

Options:

1. Continue working (don't finish session yet)
2. Mark these as deferred (update session log with reason)
3. Finish anyway (not recommended)

What would you like to do?

Wait for user decision.

### Step 3: Run Quality Checks (From Config)

**Read quality check commands from `.claude/project-config.md`** and execute in order:

1. **Format** (if configured):
- Run format command
- Report: "Format: Passed" or show errors

2. **Build** (if configured):
- Run build command
- If build fails, STOP: "Build failed. Please fix errors before finishing session."
- Report: "Build: Passed (Xs)"

3. **Lint** (if configured):
- Run lint command
- Count warnings (if any)
- Report: "Lint: Passed (N warnings)"

4. **Test** (if configured):
- Run test command
- If tests fail, STOP: "Tests failed. Please fix before finishing session."
- Report: "Tests: Passed"

5. **Custom Validation** (if configured):
- Run any custom validation commands
- Report results

**If any critical check fails (build, test), STOP and inform user.**

### Step 4: Update Session Log

1. Update "Build Status" section with results:
```markdown
## Build Status

- [x] Format: Passed
- [x] Build: Passed (12.5s)
- [x] Lint: Passed (3 warnings)
- [x] Tests: Passed
- [x] Custom: Passed
- [ ] Manual testing: [User to confirm]

2. Update "Status" at top from "In Progress" to "Complete"
3. Add completion timestamp:

   **Date**: YYYY-MM-DD
   **Status**: Complete
   **Completed**: YYYY-MM-DD HH:MM

4. If "What Was Accomplished" is empty, prompt user:

   Session log needs accomplishments summary. Please provide:
   - What features were implemented?
   - What bugs were fixed?
   - What was learned?

Step 5: Ask About Manual Testing

Ask user:

Have you tested the changes manually?
- If YES: Continue to next step
- If NO: "Please test before finishing."

Do NOT proceed until user confirms testing (or explicitly skips).

Step 6: Update Work Stream Document (if exists)

1. Check if work stream document exists: [work-streams-path]/[topic-name].md (from config)
2. If yes:
   • Add completed session to "Completed" section
   • Update status if applicable
   • Link to session log

Step 7: Update Development Log (if exists)

1. Check if development log exists at [dev-log-path] (from config)
2. If yes:
   • Add entry for this session
   • Update "Active Work" section if applicable

Suggested entry format:

## YYYY-MM-DD

### [Topic Name] Session: [Brief Title]

**Session**: [link to session log]

**Accomplished**:
- [Key accomplishment 1]
- [Key accomplishment 2]
- [Key accomplishment 3]

Step 8: Prompt to Commit

Inform the user that the session is complete and ready to commit:

Session complete!

Quality checks passed. Documentation updated.
Use /git-commit when you're ready to commit your changes.

Do NOT stage or commit files — the /git-commit skill handles that.

Step 9: Clean Up TodoWrite

1. Mark all todos as "completed"
2. Or clear the todo list entirely
3. Inform user: "Todo list cleared. Use /session-start for next session."

Project Config Template

When creating .claude/project-config.md, use this template:

# Project Configuration for Claude Skills

Project-specific settings used by Claude skills.

## Quality Checks

Commands to run before completing a session:

### Format
```bash
# Format command (e.g., cargo fmt, npm run format, go fmt)
[YOUR_FORMAT_COMMAND]

Build

# Build command (e.g., cargo build, npm run build)
[YOUR_BUILD_COMMAND]

Lint

# Lint command (e.g., cargo clippy, npm run lint)
[YOUR_LINT_COMMAND]

Test

# Test command (e.g., cargo test, npm test)
[YOUR_TEST_COMMAND]

Custom Validation (Optional)

# Any project-specific validation (e.g., RON validation, schema checks)
# Remove this section if not needed
[YOUR_CUSTOM_COMMAND]

Insight Tags

Common tags for insight-capture skill:

• architecture
• performance
• [YOUR_PROJECT_TAGS]

Documentation Paths

• Development log: docs/development-log.md
• Work streams: docs/work-streams/
• Sessions: docs/sessions/
• Decisions: docs/decisions/
• Insights: docs/insights/
• Discovery: docs/discovery/

---

## Examples

### Example 1: Clean Session Completion (Rust Project)

**User**: "finish session"

**Skill Actions**:
1. Reads `.claude/project-config.md` - finds Rust commands
2. Finds session: `[sessions-path]/user-auth/2026-01-11-auth-middleware.md`
3. Checks tasks: All 5 tasks completed
4. Runs quality checks from config:
   - `cargo fmt --all` - passed
   - `cargo build --workspace` - passed (12.5s)
   - `cargo clippy --workspace --all-targets` - passed (2 warnings)
   - `cargo test --workspace` - passed
5. Updates session log build status
6. Changes status to "Complete"
7. Asks: "Have you tested manually?" -> User: "Yes"
8. Updates work stream doc, dev log
9. Prompts: "Session complete! Use `/git-commit` to commit."

### Example 2: Build Fails

**User**: "let's wrap up"

**Skill Actions**:
1. Reads config
2. Runs format - passed
3. Runs build -> **ERROR**
4. Stops immediately
5. Reports:

Build failed! Cannot finish session with errors.

Error: [error message]

Please fix the errors before finishing the session.

6. Does NOT proceed with session completion

---

## Error Handling

**If no config and user cancels creation:**
- Warn: "Cannot run quality checks without config."
- Ask: "Skip quality checks and proceed anyway? (Not recommended)"

**If no session found:**
- Error: "No active session found. Use /session-start first?"
- List available sessions if any

**If session already complete:**
- Warn: "Session already marked complete. Finish anyway?"

**If quality checks fail:**
- STOP immediately
- Report errors to user
- Do NOT proceed until issues fixed

**If user hasn't tested:**
- STOP unless user explicitly skips
- Remind: "Manual testing is important for quality"

---

## Notes

- Quality check commands come from `.claude/project-config.md`
- If no config exists, skill prompts to create one
- Config is per-project, skills are generic
- All quality checks are optional - config defines what's needed
- Manual testing confirmation is always asked (project-agnostic)