PR Custody

Add claude-handoff

@REMvisualchecks n/achecks…add-claude-handoff → master1 files · +1 −0updated 2mo ago

▸Description· 3 comments

claude-handoff

Session handoff skills for Claude Code that capture decisions, failed approaches, measurements, and next steps — so your next session picks up exactly where you left off.

Why it exists

Claude already tries to hand off context between sessions — generating summaries with "What's broken" and "Next session paste-prompt." It looks like a handoff. It isn't one. We ran a controlled A/B test (same bug, same codebase, three fresh sessions) and found:

	Session summary (no skill)	`/handoff` skill
User had to intervene	Yes	No — self-gated
Root cause precision	Surface-level	Precise call chain
Fix correctness	Correct but unjustified	Correct and justified
Chain awareness	Zero	Referenced prior session's analysis

Full comparison data: docs/COMPARISON

Features (v1.5.0)

Deep context mining — 12-item extraction checklist with tiered strategy for large contexts (map-reduce at 500K+ tokens)
Chain tracking — handoffs link across sessions via task IDs, with sequence numbers and parent references
Self-validation — enforces line minimums and data completeness; re-mines if too thin
Anti-shadowing guard — prevents Claude from generating inferior freeform handoffs without the skill
Onboarding protocol — next session must prove comprehension before executing
PreCompact hook — safety net before context compression
Works standalone — optional integrations with Beads (task tracking) and OpenViking (AI memory)

Install

git clone https://github.com/REMvisual/claude-handoff.git
cp -r claude-handoff/skills/handoff ~/.claude/skills/
cp -r claude-handoff/skills/handoffplan ~/.claude/skills/

@REMvisual2mo ago

Both fixed in the second commit (c5b4c22) — moved to Development & Code Tools in alphabetical order and fixed the typo. Should be good to go now.

@REMvisual2mo ago

Updated the PR description with v1.5.0 features and A/B test data. The skill now includes an anti-shadowing guard (prevents Claude from generating inferior freeform handoffs) and an expanded onboarding protocol — both validated with controlled testing against Claude's built-in session summaries.

@REMvisual2mo ago

Hi! The requested changes from the review were addressed in commit c5b4c22 (moved to correct section, fixed typo). Just checking if this is good to merge now. Thanks!

loading diff…