Best practices for session documentation - incremental summaries, fix reports, and audit trails
npx claudepluginhub joshuarweaver/cascade-ai-ml-engineering --plugin delphine-l-claude-globalThis skill is limited to using the following tools:
Best practices for creating effective documentation during Claude Code sessions.
Creates isolated Git worktrees for feature branches with prioritized directory selection, gitignore safety checks, auto project setup for Node/Python/Rust/Go, and baseline verification.
Executes implementation plans in current session by dispatching fresh subagents per independent task, with two-stage reviews: spec compliance then code quality.
Dispatches parallel agents to independently tackle 2+ tasks like separate test failures or subsystems without shared state or dependencies.
Best practices for creating effective documentation during Claude Code sessions.
When working through multiple related issues in a session:
Create summary after each major fix:
FIX_SUMMARY.md - First issue resolvedFISH_CLASSIFICATION_FIX_SUMMARY.md - Second issueTAXONOMY_COMPLETE_SUMMARY.md - Final stateBenefits:
Include in each summary:
# [Issue Name] - [STATUS]
## Problem
[Clear description of what was wrong]
## Root Cause
[Why the problem occurred]
## Items Fixed (N total)
| Item | Details | Category |
|------|---------|----------|
| ... | ... | ... |
## Solution Applied
1. [Step 1]
2. [Step 2]
...
## Verification Results
✓ [What was verified]
✓ [Counts match]
## Files Modified
1. `file1.txt` - [what changed]
2. `file2.txt` - [what changed]
## Next Steps
[What the user should do now]
Use descriptive names that indicate:
FISH_CLASSIFICATION_FIX_SUMMARY.mdTAXONOMY_COMPLETE_SUMMARY.mdAvoid generic names like SUMMARY.md or NOTES.md.