Step 1 of 4
Define the three schemas
Each memory kind answers a different question - schema follows.
Define schemas for the three stores. EPISODIC (what happened): {session_id, date, task, outcome, notable_events[], artifacts_touched[]} - compact narrative records of sessions and incidents. SEMANTIC (what is true): {statement, topic, provenance, confidence, source_episodes[]} - atomic declarative facts. PROCEDURAL (how to do): {goal, preconditions[], steps[], verification, failure_modes[], last_validated, source_episodes[]} - executable recipes. Write two real examples per store from my project history, including one entry that must be SPLIT (a session story containing both a new fact and a new recipe) to demonstrate the routing discipline.
Expected after this step
Three schemas with real examples including a demonstrated split.
Should not happen
- ✕Facts phrased as stories, so recall returns anecdotes when the agent needs truth
- ✕Procedures without verification steps - recipes that cannot be trusted
- ✕Episodes hoarded forever at full fidelity, bloating the store
- ✕No links between layers, so a failed episode never corrects the stale procedure
Verify before continuing
Do not move on until every check is true. The complete button stays locked until then.
Do not continue if…
- !Facts phrased as stories, so recall returns anecdotes when the agent needs truth
- !Procedures without verification steps - recipes that cannot be trusted
- !Episodes hoarded forever at full fidelity, bloating the store
- !No links between layers, so a failed episode never corrects the stale procedure
If the AI messes this up
Use this when the AI fakes progress or breaks the feature. It forces a real fix.
The examples are interchangeable across stores. Sharpen with the question test: episodes answer 'what happened?', facts answer 'what is true?', procedures answer 'how do I?' - rewrite each example until it answers only its own question.