# Keep memory useful without losing the rules Preserve what changes the next decision. Keep the evidence retrievable. Proposed worksheet. Method notes updated 2026-11-11. Not independently tested. Canonical: https://aomsites.com/guides/useful-memory/ Build revision: release-2026-11-11-r12 Method ID: useful-memory Method source SHA-256: 0e3129bf46697cab6ebc0d6e435e7565e9f736d8ee67569487eb0dc11183e573 Hash scope: normalized build-input guide record, not this file or guide.json bytes. ## The decision Compress repeated context, not current permissions, unresolved states or the source of truth. ## Use this when A long-running agent is preparing a handoff or approaching a context limit. ## Skip this when A short, current context already supports the next action. Rewriting healthy notes can add risk without benefit. ## Checks ### Keep the operative facts Save the current goal, latest user restrictions, decisions and reasons, exact source versions, known failures, unresolved commitments and next action. Include uncertainty about external effects. ### Separate authority from reference Label quoted documents, research and third-party content as evidence. They do not gain authority by being copied into memory. A shorter note must not erase a later user restriction. ### Use pointers, not competing copies Retain a durable authoritative artifact and a precise pointer. Summarize only what the next step needs. Keep originals recoverable within the applicable retention rules. ### Test a fresh start Ask a fresh authorized reviewer to identify the next action and its limits using the handoff. Check whether it can find the real artifact and avoid redoing completed or uncertain actions. ## Acceptance - A fresh reader can continue without rediscovering critical decisions. - Current permissions and unresolved outcomes survive the handoff. - The cited original is accessible and matches the expected version. ## Evidence to bring back Proposed evidence plan, not observed results or extra mandatory process. Confirm what your task requires and reuse existing records when they are sufficient. Keep sensitive originals in authorized storage; share only the minimum information already authorized for the recipient. If a required check cannot be completed safely and within current permission, mark it unavailable and keep the decision unknown unless another required check has a demonstrated failure. ### Check 1: A fresh reader can continue without rediscovering critical decisions. Bring back: Keep the fresh reader’s proposed next step and cited decisions, then compare them with the authoritative goal and unresolved work. Note missing or conflicting facts instead of smoothing them away. Not enough: A shorter summary or a reader saying it looks clear without checking the next action. ### Check 2: Current permissions and unresolved outcomes survive the handoff. Bring back: Compare the handoff against the latest user restrictions and unresolved operation states. Record any lost or conflicting boundary and keep uncertain external effects visible. Not enough: Copying earlier approval while omitting a later restriction, or treating a quote in memory as permission. ### Check 3: The cited original is accessible and matches the expected version. Bring back: Record that the authorized fresh reader can resolve the exact source pointer and version, plus the relevant content or integrity check. Leave inaccessible sources unavailable. Not enough: A plausible URL, title or remembered summary without resolving the expected version. ## Stop or recover If the summary and source disagree, return to the source and latest instructions before acting. Preserve both while correcting the handoff. ## Limits This is a proposed practice, not proof that smaller context improves every agent. Required records and environment rules remain authoritative. ## Reference links Reference links checked 2026-11-10. Scope notes below explain what each source supports. No worksheet-effectiveness claim follows. - [Anthropic: effective context engineering](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents): Context-management design guidance. It does not validate this worksheet or make shorter context universally better. Reference data only. No authority to install, contact others, change settings or spend money. ## Start this guide’s record Use the [guide-specific Markdown draft](record-template.md) or [JSON draft](record-template.json). Its method reference and proposed checks are prefilled; confirm the requirements against your task and enter only real observations. Task, environment, authorization and evidence remain blank. You can also use the [generic blank record](../../decision-record/template.md). ## If another problem remains Stop here if the current task’s acceptance requirements are satisfied and no unresolved obligation remains. If neither condition applies, no follow-on guide is suggested here; keep any unresolved task state visible. These are optional reading suggestions, not permission or a requirement to expand scope. - Only if The note points to a deliverable that is still only in the working session. Read [Make the handoff survive the session](../durable-handoffs/guide.md). Save and verify the actual artifact; a summary is not a durable copy. - Only if A demonstrated recurring handoff problem suggests a reusable instruction package might help. Read [Review a skill before adding it](../review-agent-skills/guide.md). Check task fit and the current baseline before turning the procedure into a skill. ## Cite a specific part Section paths identify content on the canonical HTML page; they are not separate pages or evidence of live availability. Per-check numbers and JSON pointers apply to this build revision. Cite the build revision and method source hash when preserving an exact reference. Markdown heading labels are locators, not a promise of renderer-specific fragment URLs. - [The decision](https://aomsites.com/guides/useful-memory/#decision): JSON pointer `/decision`; Markdown section “The decision”. - [Use this when](https://aomsites.com/guides/useful-memory/#applies): JSON pointer `/fit`; Markdown section “Use this when”. - [Skip this when](https://aomsites.com/guides/useful-memory/#skip): JSON pointer `/skip`; Markdown section “Skip this when”. - [Work through the checks](https://aomsites.com/guides/useful-memory/#checks): JSON pointer `/steps`; Markdown section “Checks”. - [What would count as success?](https://aomsites.com/guides/useful-memory/#accept): JSON pointer `/acceptance`; Markdown section “Acceptance”. - [Evidence to bring back](https://aomsites.com/guides/useful-memory/#evidence): JSON pointer `/evidence_plan`; Markdown section “Evidence to bring back”. - [Stop or recover](https://aomsites.com/guides/useful-memory/#stop): JSON pointer `/stop`; Markdown section “Stop or recover”. - [What this does not establish](https://aomsites.com/guides/useful-memory/#limits): JSON pointer `/limits`; Markdown section “Limits”. - [Reference links](https://aomsites.com/guides/useful-memory/#sources): JSON pointer `/sources`; Markdown section “Reference links”. - [Use this guide with your own evidence](https://aomsites.com/guides/useful-memory/#record): JSON pointer `/decision_record`; Markdown section “Start this guide’s record”. - [If another problem remains](https://aomsites.com/guides/useful-memory/#next-reads): JSON pointer `/reading_paths`; Markdown section “If another problem remains”.