Back to catalog / Agent workflows
Carry approved working memory into the next run
Persist reusable project facts outside the conversation and retrieve only relevant notes on a later run.
Goal
Persist reusable project facts outside the conversation and retrieve only relevant notes on a later run.
Steps
- Choose a project-scoped memory directory and store short facts with source, date, and confidence instead of full transcripts. [source06]
- Before a new task, search memory for relevant notes and include only the selected excerpts in the prompt context. [source06]
- Review writes for secrets and stale facts; test that a new session can retrieve an approved note without exposing another project’s memory. [source06]
How to verify
In a fresh run, retrieve one approved note for its project and confirm a second project’s note and any secret-like fixture are excluded.
Troubleshooting
If retrieval returns stale or cross-project facts, add project scoping and date/confidence filtering; remove secrets from stored notes before indexing.
Permissions
Prerequisites
- Python 3.10+, Anthropic API key, and a private storage location with project-level access controls.
Permissions
- Use approved test or anonymized data only; scope credentials to each workflow, and require explicit approval for external writes.
Limitations
Model output may be incomplete or wrong; API and beta behavior can change, so verify against the pinned source and current service behavior.
Pinned upstream source
- Repository
- anthropics/anthropic-cookbook
- File path
- tool_use/memory_cookbook.ipynb
- Pinned commit
9c1e493ddd2b82df399d2a61b9c806c539e18be6- Source checked
- 2026-10-14 · Reviewed against the pinned upstream source
Repository metrics
These values describe the upstream repository, not popularity or likes for an individual skill or playbook.
- Stars
- Not collected
- Forks
- Not collected
- Observed at
- Not collected
No metrics snapshot is available yet.
License: MIT
Main topic
Turn ecommerce campaign materials into a reviewable weekly brief