Toolkit
Persistent Memory · Pitfalls
On this page7 sections
Pitfalls
Stale State with No Expiry
Branches, incidents, deadlines, and product behavior age quickly. Store a review date and expire aggressively.
Contradictory Memories
Appending corrections leaves multiple plausible truths. Supersede or delete the old claim and preserve correction history separately.
Memory as Policy
Machine-authored recall is not a reliable home for required team rules. Keep authoritative guidance checked in or managed.
Cross-Scope Leakage
User, repository, role, and hosted account scopes can expose context to the wrong task. Verify isolation and access controls.
Inference Stored as Fact
Label observed fact, inference, assumption, and preference distinctly. Link evidence for load-bearing claims.
Hidden Storage and Retention
Operators need to know where memory lives, who can read it, how long it remains, and how to delete it.
Falsification Checklist
- What is the source and review date?
- Does the claim still match current code or documentation?
- Could another scope or AgentA worker you delegate to: brief it, and it takes steps on its own. A chatbot answers; an agent acts.An LLM that runs tools in a loop toward a goal — it acts, checks the result, decides the next step, and repeats until done.Full definition see it unexpectedly?
- Can a user inspect and correct it?
- Would the system remain safe if the memory were wrong?
Source register
Persistent Memory sources
Primary vendor documentation reviewed on the date shown. Links may change after publication.
- How Claude remembers your project (opens in a new tab)
Anthropic · official docs · accessed 2026-08-01
- Custom instructions with AGENTS.md (opens in a new tab)
OpenAI · official docs · accessed 2026-08-01
- Memories (opens in a new tab)
OpenAI · official docs · accessed 2026-08-01
- GitHub Copilot CLI command reference (opens in a new tab)
GitHub · official docs · accessed 2026-08-01