DEDUPED Hot AI repos, turned into use cases Get the Monday 5

How to hand off an unfinished task from one agent to another with Shared Memory Vault

Let a second agent resume work from a shared CONTEXT.md without re-explaining the task.

Short answer: Set up the knowledge base once with vault-init.sh so both agents share the same folder; then Have the first agent record its goal, decisions and next step in CONTEXT.md for the session; Have agents update CONTEXT.md incrementally every 3 turns: python3 ~/.workbuddy/skills/shared-memory-vault/scripts/context-update.py. All steps.

Who it is forA developer whose coding agent hits a session limit or switches device and needs a second agent to continue the same task.

The problemThe next agent starts cold, repeats earlier steps, or makes decisions that conflict with the first agent's choices.

The outcomeThe second agent reads one CONTEXT.md file and continues from where the first agent stopped, with decisions kept consistent.

What you need first

  • Python to run the standard-library scripts
  • bash to run vault-init.sh
  • A cloud-synced directory on your machine
  • An AI agent that can read SKILL.md and follow its lifecycle SOP

Steps

  1. Set up the knowledge base once with vault-init.sh so both agents share the same folder.
  2. Have the first agent record its goal, decisions and next step in CONTEXT.md for the session.
  3. Have agents update CONTEXT.md incrementally every 3 turns: python3 ~/.workbuddy/skills/shared-memory-vault/scripts/context-update.py (inferred, check the README)
  4. Start the second agent and tell it to read CONTEXT.md first, then continue the task.
  5. When the task pauses, have the agent archive the session with vault-archive.py so the next handoff starts from a clean blackboard.

Prompt for your agent

Using the shared-memory-vault repo, read the CONTEXT.md file in <knowledge base path>/context/ and resume the task it describes. Report the next step you take and the decision you are following from the file. Verify that you did not change any earlier decision without noting it.

Caveat. The README describes the blackboard convention but does not define the exact CONTEXT.md fields, so agree on a format with your agents before the first handoff.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for songs-aaa/shared-memory-vault and "Hand off an unfinished task from one agent to another"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer