Skip to content

feat: add agent-handoff skill + autoMemoryEnabled doc note - #8

Open
mickzijdel wants to merge 1 commit into
mainfrom
claude/youthful-noether-n5yqti
Open

mickzijdel wants to merge 1 commit into
mainfrom
claude/youthful-noether-n5yqti

Conversation

@mickzijdel

Copy link
Copy Markdown
Owner

Summary

Two improvements from this commit-digest run (2026-06-13):

1. New agent-handoff skill (thinking-tools, minor bump → 1.2.0)

Inspired by feed item 2026-06-10: "Coordinate two agents via a shared plan.md: one writes the plan, the other loops on reading it, executing todos, and reporting updates back in the file."

The skill documents the plan.md coordination pattern for multi-session and multi-agent work. It pairs with the existing plan-reminder.sh Stop hook in dev-hooks (which already fires when .claude/current_plan.md exists but is unchanged at session end).

Covers:

  • When to use plan.md (continuing work, parallel agents, multi-day projects)
  • Canonical plan.md format (Goal as verification statement, Context, Tasks with checkboxes, Status)
  • How to start a session from a plan (read Status first, scan done tasks, begin from first [ ])
  • Parallel agent coordination pattern (planner agent A + executor agent B)
  • Handoff note format (current state, blockers, exact next action)
  • Connection to plan-reminder.sh
  • Tips (Goal as verification, mid-session Status updates, lean Context, coarse tasks)

2. autoMemoryEnabled: false recommendation in dev-hooks README (dev-hooks, patch bump → 2.2.1)

Nate Berkopec disabled Claude Code auto-memory in dotfiles commit 609e48c (2026-06-10). The same reasoning applies here: memory-reminder.sh already owns structured, intentional end-of-session memory capture. With auto-memory enabled (the default), Claude may write to memory automatically during sessions, producing redundant entries. The note recommends setting "autoMemoryEnabled": false in ~/.claude/settings.json so all memory writes go through the end-of-session review hook.

Test plan

  • Read agent-handoff/SKILL.md — does the format section match how you'd actually use plan.md?
  • Check that the skill table row in thinking-tools/README.md reads cleanly
  • Review the autoMemoryEnabled note in the dev-hooks README memory-reminder bullet

Generated by Claude Code

…memory-reminder

agent-handoff (thinking-tools, minor bump → 1.2.0): a plan.md coordination skill
for multi-session and multi-agent work. Pairs with the existing plan-reminder.sh
Stop hook. Inspired by feed item 2026-06-10 on coordinating agents via shared plan.md.

memory-reminder autoMemoryEnabled note (dev-hooks, patch bump → 2.2.1): adds a
sentence to the memory-reminder README note recommending autoMemoryEnabled: false
in global Claude settings. Nate disabled it in dotfiles commit 609e48c for the same
reason — the hook already owns intentional end-of-session memory capture.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant