Harvest Memory
Productivityharvest-memoryHarvest Memory is an Agent Skill for productivity published in the shopsys/shopsys repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Extracts key learnings from a session and persists them using Memory MCP tools.
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/shopsys/shopsys/blob/HEAD/.agents/skills/harvest-memory/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/harvest-memory, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Harvest Memory
You are tasked with reviewing the FULL conversation transcript now in context and extract:
- Topic / objective of the session
- Approaches/strategies attempted
- Dead ends / what went wrong
- Prompts or actions that led to good outcomes
- Architecture discoveries about the app (esp. "storefront") and any other systems
- Future improvements / refactors / optimizations
Then use the Memory MCP tools to persist this into the knowledge-graph memory (do NOT handcraft the structure yourself):
-
Ensure entities exist:
- A session entity (name it with a stable ID, e.g.,
Session_<ISO date>_<short topic>) of type"conversation". - The relevant project entity (e.g.,
"StorefrontApp") of type"project", creating it if missing. Usecreate_entities.
- A session entity (name it with a stable ID, e.g.,
-
Relate entities:
- Link the session → project with an active-voice relation (e.g.,
"concerns"). Usecreate_relations.
- Link the session → project with an active-voice relation (e.g.,
-
Add observations:
- For the session entity, add granular observations for topic, approaches, dead ends, successes, and any notable prompts.
- For the project entity, add architecture findings and future improvements.
Use
add_observations.
Each observation must be a single fact (atomic). If you’re unsure which entity to attach something to, prefer the project entity for architecture/improvement facts and the session entity for process/history facts.
Execute the memory tool calls now so the data is actually stored.
Ultrathink.