memory

Umbrella entrypoint for Memory Layer workflows; use for broad memory-related turns, shared invariants, and docs/admin work, while focused query/resume, plan-execution, and remember skills handle the narrow workflows

Record a task memory before starting direct no-plan implementation work when the user asks to implement, fix, change, add, update, release, install, or otherwise mutate the repository without first approving a plan

Guide first-time GitHub repository initialization for a Memory Layer project; use when setting up or auditing GitHub remotes, repository metadata, Actions workflows, secrets, variables, branch protection, required checks, or Memory Layer repo onboarding, with discovery and dry-run review before write-capable changes

Save the approved plan and checkpoint when execution starts, and verify all plan checkbox items are complete before claiming the task is finished

Initialise or refresh Memory Layer in a project directory after Memory Layer is already installed; use for repo-local setup with .mem, .agents/memory-layer.toml, and repo-local memory skills, previewing writes before applying them

Query curated project memory before answering repo-specific questions, and use resume to get back into flow after interruptions

Remember meaningful completed work by capturing task context and curating it into durable project memory with provenance

Review pending Memory Layer replacement proposals interactively inside the LLM CLI; explain why each memory update was proposed, gather codebase proof when needed, and approve or reject only after explicit confirmation

Query project memory before answering project-specific questions; capture completed task context; curate raw captures into durable canonical memory with provenance.